1,348 248 2MB
Pages 302 Page size 422.88 x 638.4 pts Year 2010
Stalking the Riemann Hypothesis The Quest to Find the Hidden Law of Prime Numbers Dan Rockmore
Pantheon Books
-
11111111
New York
Copyright © 2005 by Daniel N. Rockmore All rights reserved under International and Pan-American Copyright Conventions. Published in the United States by Pantheon Books, a division of Random House, Inc., New York, and simultaneously in Canada by Random House of Canada Limited, Toronto. Pantheon Books and colophon are registered trademarks of Random House, Inc. Library of Congress Cataloging-in-Publication Data Rockmore, Daniel N. (Daniel Nahum) Stalking the Riemann hypothesis : the quest to find the hidden law of prime numbers/Dan Rockmore. p. cm. Includes index. ISBN 0-375-421 36-X 1. Numbers, Prime. 2. Number theory. 3. Riemann, Bernhard, 1 826-1 866. 1. Title. QA246.R63 2005 51 2.7'23--dc22 2004053525 www.pantheonbooks.com Printed in the United States of America First Edition 1 234 5 6 7 8 9
For Loren Butler Feffer (1962-2003), scholar, athlete, and dearfriend
CONTENTS
Acknowledgments
2 3 4 5 6 7 8 9 10 II I2 13 14 15
ix
Prologue-It All Begins with Zero 3 The God-Given Natural Numbers 1 The Shape ofthe Primes 21 Primal Cartographers 30 Shoulders to Stand Upon 46 Riemann and His "Very Likely" Hypothesis 63 A Dutch Red Herring 95 A Prime Number Theorem, After All... and More 105 Good, but Not Good Enough 118 First Steps 128 A Chance Meeting ofTwo Minds 154 God Created the Natural Numbers ... but, in a Billiard Hall? 186 Making Order Out of(Quantum) Chaos 213 God May Not Play Dice, but What about Cards? 232 The Millennium Meeting 261 Glossary 261 Further Reading and Sources 211 Index 219
ACKNOWLEDGMENTS
T H E SCIENCE WRITER faces an interesting intellectual challenge. To distill years, even centuries, of scientific investigation for a broad and curious audience, while not raising the hackles of the experts in the field, is something of an intellectual tightrope walk. It is my hope that in this book I haven't fallen off the wire too many times, and I'm grateful to the family, friends, and colleagues who have either steadied me during this performance, or at least allowed me a soft landing from time to time. I'd like to take a moment to thank some of them. I have been lucky to have had comfortable circumstances in which to write, and I appreciate the hospitality provided to me by New York University's Courant Institute, the Institute for Advanced Study, the Santa Fe Institute, and of course my home turf: Dartmouth College's departments of mathematics and computer science. I would like to give special thanks to the Institute for Advanced Study's librarian Momota Ganguli for her help in tracking down a few hard-to-find references, as well as Dartmouth College's librarian Joy Weale for the quick attention she gave to my many requests for materials. Katinka Matson helped me find a home for this book with Ran dom House, where Marty Asher, Edward Kastenmeier, Will Sulkin, and Eric Martinez have worked hard to steer it through the shoals of the editorial and production process. Special thanks to Susan Gamer for a terrific job of copyediting. I am grateful to the many people who read portions of the book in various stages of completion. In particular, I'd like to thank Percy
x
Acknowledgments
Deift, Freeman Dyson, Cormac McCarthy, Peter Sarnak, Craig Tracy, Harold Widom, and Peter Woit for their helpful comments and cor rections, Eric Heller and Jos Leys for furnishing some key figures, and Andrew Odlyzko for providing easy access to the zeta zero data. Sin cere thanks to Peter Doyle for translating the relevant pieces of the Stieltjes-Hermite correspondence, and for keeping local interest alive through his leadership of the always entertaining "Riemann Seminar" at Dartmouth. Peter Kostelec, a terrific scientist and friend, made many of the figures in this book, as well as many more that weren't used. Bob Drake, resident "folk mathematician," gave me many good comments on early versions of the manuscript and many more good squash lessons. Thanks to Laurie Snell for his close reading of an early manuscript and for his constant encouragement. Greg Leibon's care ful attention helped me avoid a few bloopers at the end. Of course, I take full responsibility for any errors that made it past the eagle eyes of my vigilant friends and colleagues. My parents, Ron and Miriam; my brother Adam; my sister-in-law Alicia; and my delightful niece Lucy always have given me the luxury and benefit of a loving and supportive family. If it is true that the apple doesn't fall too far from the tree, I am lucky to have come from sturdy stock giving generous shade. Ellen-my wife, love, confidante, and super-editor-and our son, Alex, have brought to my life a light and joy that I had not thought possible. I'm grateful to them every day. Through every stage of this project, my dear and now departed dog Digger was a source of con stant comfort and calm. I always appreciated and now miss his nuz zling reminders of the necessity of regular breaks-in good weather and bad. If there is a sonorous sentence or two in this book, it was almost surely rocked loose by the steady footfall of one of our count less but countable walks. I like to think that his playful and gentle spirit lives on in these pages. Finally, this book is dedicated to the memory of my dear friend Loren Butler Feffer, whose keen mind, boisterous good humor, and fierce loyalty helped me through many a difficult time, both personal and professional. She is ever missed. Loren-I wish you could have been there when we popped open the champagne. This one's for you.
STALKING THE RIEMANN HYPOTHESIS
Prologue-It All Begins with Zero
IT's ONE OF THOSE
slate-gray summer days that more properly belong to mid-August than late May, one of those days in New York City when it is barely clear where the city ends and the sky begins. The hard-edged lines and Euclidean-inspired shapes that are building, sidewalk, and pavement all seem to fuse into one huge melted mass that slowly dissolves into the humid, breezeless, torpid air. On mornings like this, even this irrepressible metropolis seems to have slowed a notch, a muffled cacophony more bass than treble, as the city that never sleeps stumbles and shuffles to work. But here in Greenwich Village, at the corner of Mercer and West Fourth streets, where we find New York University's Warren Weaver Hall, the hazy torpor is interrupted by a localized high-energy eddy. Here, deep in the heart of the artistic rain forest that is "the Village," just across the street from the rock 'n' rolling nightclub the Bottom Line, a stone's throw from the lofts and galleries that gave birth to Jackson Pollock, Andy Warhol, and the Velvet Underground, is the home of the Courant Institute of Mathematical Sciences, where at this moment there is an excitement worthy of any gallery opening in SoHo, or any new wave, next wave, or crest-of-the-wave musical performance. The lobby and adjacent plaza are teeming with mathematicians, a polyglot and international group, abuzz with excitement. Listen closely, and amid the multilingual, every-accent mathematical jibber jabber you'll hear a lot of talk about nothing, or more properly a lot of talk about zero.
4
STALK I N G T H E RIEMANN H YPOT H E S I S
Zero is not an uncommon topic of conversation in New York, but more often than not it's the "placeholder zeros" that are on the tip of the New Yorker's tongue. These are the zeros that stand in for the orders of magnitude by which we measure the intellectual, cultural, and financial abundance that is New York: one zero to mark the tens of ethnic neighborhoods, two for the hundreds of entertainment options, three for the thousands of restaurants, six for the millions of people, and, of course, the zeros upon zeros that mark the billions or even trillions of dollars that churn through the city every day. These are not the zeros of void, but the zeros of plenty. But, today, just one week past Memorial Day 2002, it's a zero of a different flavor which has attracted this eclectic group to downtown New York City. Here some of the world's greatest mathematicians are meeting to discuss and possibly, just possibly, witness the resolution of the most important unsolved problem in mathematics, a problem that holds the key to understanding the basic mathematical elements that are the prime numbers. The zeros that tip the tongues of these mathematical adventurers are zeta zeros, * and the air is electric with the feeling that perhaps this will be the day when we lay to rest the mystery of these zeros, which constitutes the Riemann hypothesis. For over a century mathematicians have been trying to prove the Riemann hypothesis: that is, to settle once and for all a gently asserted conjecture of Bernhard Riemann ( 1 826-1 866), who was a professor of mathematics at the University of Gottingen in Germany. Riemann is perhaps best known as the mathematician responsible for inventing the geometrical ideas upon which Einstein built his theory of general relativity. But in 1 859, for one brief moment, Riemann turned his attention to a study of the long-familiar prime numbers. These are numbers like two, three, five, and seven, each divisible only by one and itself, fundamental numerical elements characterized by their irreducibility. Riemann took up the age-old problem of trying to find a rule which would explain the way in which prime numbers are dis tributed among the whole numbers, indivisible stars scattered with out end throughout a boundless numerical universe. In a terse eight-page "memoir" delivered upon the occasion of his *Terms in
boldface italic can be found in the glossary.
Prologue-It All Begins with Zero
5
induction into the prestigious Berlin Academy, Riemann would revo lutionize the way in which future mathematicians would henceforth study the primes. He did this by connecting a law of the primes to the understanding of a seemingly completely unrelated complex collec tion of numbers-numbers characterized by their common behavior under a sequence of mathematical transformations that add up to the Riemann zeta function. Like a Rube Goldbergesque piece of mathe matical machinery, Riemann's zeta function takes in a number as raw material and subjects it to a complicated sequence of mathematical operations that results in the production of a new number. The rela tion of input to output for Riemann's zeta function is one of the most studied processes in all of mathematics. This attention is largely due to Riemann's surprising and mysterious discovery that the numbers which seem to hold the key to understanding the primes are precisely the somethings which Riemann's zeta function turns into nothing, those inputs into Riemann's number cruncher that cause the produc tion of the number zero. These are the zeta zeros, or more precisely the zeros of Riemann's zeta function, and they are the zeros that have attracted a stellar cast of mathematicians to New York. In his memoir, Riemann had included, almost as an aside, that it seemed "highly likely" that the zeta zeros have a particularly beautiful and simple geometric description. This offhand remark, born of genius and supported by experiment, is the Riemann hypothesis. It exchanges the confused jumble of the primes for the clarity of geome try, by proposing that a graphical description of the accumulation of the primes has a beautiful and surprisingly simple and precise shape. The resolution of Riemann's hypothesis holds a final key to our understanding of the primes. We'll never know if Riemann had in mind a proof for this asser tion. Soon after his brief moment of public glory, the ravages of tuber culosis began to take their toll on his health, leaving him too weak to work with the intensity necessary to tie up the loose ends of his Berlin memoir. Just eight years later, at the all too young age of thirty-nine, Riemann was dead, cheated of the opportunity to settle his conjecture. Since then, this puzzling piece of Riemann's legacy has stumped the greatest mathematical minds, but in recent years frustration has begun to give way to excitement, for the pursuit of the Riemann
6
STALK I N G T H E RIEMANN H YPOT H E S I S
hypothesis has begun to reveal astounding connections among nuclear physics, chaos, and number theory. This unforeseen conflu ence of mathematics and physics, as well as certainty and uncertainty, is creating a frenzy of activity that suggests that after almost 1 50 years, the hunt might be over. This is the source of the buzz filling the Courant Institute's entry way. It is a buzz amplified by the fact that whoever settles the question of the zeta zeros can expect to acquire several new zeros of his or her own, in the form of a reward offered by the Clay Institute of Mathe matics, which has included the Riemann hypothesis as one of seven "Millennium Prize Problems," each worth $ 1 million. So the jungle of abstractions that is mathematics is now full of hungry hunters. They are out stalking big game-the resolution of the Riemann hypothe sis-and it seems to be in their sights. The Riemann hypothesis stands in relation to modern mathemat ics as New York City stands to the modern world, a crossroads and nexus for many leading figures and concepts, rich in unexpected and serendipitous conjunctions. The story of the quest to settle the Rie mann hypothesis is one of scientific exploration and discovery. It is peopled with starry-eyed dreamers and moody aesthetes, gregarious cheerleaders and solitary hermits, cool calculators and wild-eyed visionaries. It crisscrosses the Western world and includes Nobel lau reates and Fields medalists. It has similarities with other great scien tific journeys but also has its own singular hallmarks, peculiar to the fascinating world of mathematics, a subject that has intrigued humankind since the beginning of thought.
~
The God-Given Natural Numbers
THE GREAT GERMAN mathematician Leopold Kronecker (1 823-1 89 1 ) said that "God created the natural numbers." And it is true that the natural nu mbers-one, two, three, four; on and on they go-appear to have been present from the beginning, coming into existence with the birth of the universe, part and parcel of the original material from which was knit the ever-expanding continuum of space-time. The natural numbers are implicit in the journey of life, which is a nesting of cycles imposed upon cycles, wheels within wheels. One is the instant. Two is the breathing in and out of our lungs, or the beat of our hearts. The moon waxes and wanes; the tides ebb and Row. Day follows night, which in turn is followed once again by day. The cycle of sunrise, noon, and sunset give us three. Four describes the circle of seasons. These natural numbers help us to make sense of the world by finding order, in this case an order of temporal patterns, that lets us know what to expect and when. We notice the rising and setting of the sun, and that cycle of two is given a more detailed structure as we follow the sun through the sky over the course of a day. We turn the temporal telescope around and also see day as part of the larger cycle of the phases of the moon, whose steady progress is situated within the cycle of the seasons that makes up the year. Patterns within patterns within patterns; numbers within numbers within numbers-all working together to create a celestial symphony of time.
8
STALK I N G T H E RIEMANN H YPOT H E S I S
Armed with this new understanding we make tentative, tiny forays out into the Jamesian "booming, buzzing world" and shape a life within and around it. Embedded in the recognition of the cycle is the ability to predict, and thus to prepare, and then to direct the world to our advantage. We coax and bend an unflinching, steady march of time; and in a subtle jujitsu of nature, technologies are born. We learn when to sow and when to reap, when to hunt and when to huddle. We exploit that which we cannot change. We discover the cycle and ride it as an eagle rides an updraft. In the absence of a natural cycle we may impose one, for in routine we find a sense of control over the unwieldy mess that is life. We relish the comfort of being a regular at a local diner or a familiar face at the coffee cart on the street, and the rhythm of the daily morning dog walk. We dream of options, if only to choose our own routines, our own patterns, our own numbers. But as befits that which is part and parcel of space-time, number is not only a synecdoche of temporal organization but also the most basic and elementary means of quantifYing a spatial organization of the world. Nature gives us few, if any, truly straight lines or perfect cir cles. But there is one moon; there is one sun; the animals go two by two. We organize, we count, and therefore we are. In this way, number is presented to us in the world in both time and space, instances upon instances, but this is only the beginning. Kronecker said not only that the natural numbers are God-given, but also that "all else is the work of man." What first appear as singular phenomena are eventually unified, gathered into a collective that is then recognized as a pattern. Soon, the pattern is itself familiar, and so it becomes less a pattern and more a particular. The game is then repeated, and we find a new superpattern to explain what had once seemed disparate patterns. So on and so on we go, building the disci pline that will come to be known as mathematics. BEGINNING "THE WORK"
Suppose that I walk past a restaurant and catch a glimpse of a per fectly set square table, place settings at each edge, each side of the table providing a resting place for a full complement of plates, glasses,
The God-Given Natural Numbers
9
and silverware. As I approach the entrance to the restaurant, a group of women arrive and I imagine them seated at that beautiful table, one at each side, continuing their animated conversation. As I pass by again some time later, I see the women leave the restaurant. They stand outside, say their good-byes, and one by one are whisked away by taxicabs. What is it that the group of women has in common with the col lection of place settings, the chairs at the table, the very sides of the table, and the taxicabs that finally take them away? It is the corre spondence that they engender. It is a correspondence that I make mentally and visually as I watch the women, one that you make as you read this story, seeing each woman paired with a chair, a plate, or a taxi. Any other grouping of objects that could be paired with them in this way has this same property, this same basic pattern. This pat tern is one of "numerosity." These groups all share the property of "fourness." Each collection, whether it be the chairs, the place settings, or the taxis, is such that its component objects can be put into a one-to-one correspondence with the group of women. We say, as an abbreviation for this property, that the group of women has a size of four, and this is a property shared among each of the sets of objects that may be put into a correspondence with the women. If you had in your possession a collection of hats and I inquired if you had one for each of the women, you might have me list the women, or show you a picture of the group, but even better, you could ask me "how many" women need hats. My answer, "four," would be enough for you to check to see if you had one hat for each. The self-contained nature of the correspondence-there is no object left unpaired-is perhaps what underlies the other classifica tion of the number four, or for that matter any natural number, as an integer, and in particular a positive integer. The totality of the integers consists of the natural numbers, their negatives, and zero. Thus four becomes an agreed-upon name for a pattern that we rec ognize in the world. At Christmas, four are the calling birds; at Passover, four are the matriarchs, symbols that are simultaneously iconic and generic. We wind our way back through numerical history. Four are the fingers proudly displayed by a protonumerate toddler, a
10
STALK I N G T H E RIEMANN H YPOT H E S I S
set of scratchings on a Sumerian cuneiform, or the bunch of beads or pebbles lying at the feet of a Greek philosopher. The last of these uni versal physical numerical proxies, which the Greeks called calculi, gave birth to our words calculus and calculate, and mark the mathe matician as both the forefather and the child of the first "bean coun ters": the Pythagoreans. The Pythagoreans, followers of the great Greek mathematician Pythagoras (580-500 B.C.E.), are said to have first abstracted the par ticular instances of a number to a universal and generic pile of pebbles or arrangements of dots in the sand. For the Pythagoreans, mathe matics and the mathematical were ubiquitous. They lived by the motto "Number is all," and as a consequence numerology was inte gral to their daily lives. As for Pythagoras, very little is known of his life. He was a mystic and scholar who led a cult whose basic tenet was devotion to the study of knowledge for its own sake. To describe this pursuit, Pythagoras coined the word philosophy. As applied to the spe cific study of number and geometry, this sort of investigative approach led him to create another discipline, which he named math ematics, meaning "that which is learned." Before the Pythagoreans, mathematical studies were by and large contextualized through their particular applications. Records are sketchy, but certainly the Egyptians and Babylonians developed a great deal of basic mathematical (or, perhaps more accurately, "calcu lational") procedures for dealing with the everyday problems of their agricultural, architectural, or commercial activities. The mathematical investigations of the Pythagoreans appear to mark the origins of a pure study of number, as opposed to a study of number in the context of its potential or actual applications. So Pythagoras, the "father of mathematics," gave birth to a poetry or art of pattern, an intellectual discipline in which numbers themselves were studied as arrangements of beautiful, regular, rhythmic shapes. Herein lies the origin of a study of the laws that govern patterns within the patterns or the behavior of numbers. This is number theory, and the Pythagoreans are history's first number theorists. The simplest patterns to be found were those of the geometric shapes in which only certain numbers of pebbles could be arranged. There were square numbers, such as one, four, nine, and sixteen,
The God-Given Natural Numbers
II
and triangular numbers such as one, three, and six,
and even hexagonal numbers like one and six,
each providing an infinite family of dimpled patterns ready for cate gorization and study. These few examples represent but a shallow draw from a potentially infinite well of possibilities. Any of your favorite two-dimensional shapes could yield other classes offigurate numbers, and many others were studied.
PRIMES: THE FUNDAMENTAL PATTERN WITH IN THE PATTERN
The Pythagoreans plumbed the piles of pebbles in search of sttucture and regularity, and so might I also look for patterns as I pass by my lit tle imagined restaurant. One evening I peer though the window and notice that the room is full. How many diners are there on this busy evening? I could count them all one by one if I stood there long enough, but I find that I can't help trying to take a shortcut. The tables organize the diners into groups of four, and so I know that the num ber of diners will simply be four times the number of tables. The tables are arranged tastefully and delicately about the softly lit room,
12
STALK I N G T H E RIEMANN H YPOT H E S I S
and again my eye looks for patterns. I first cluster the tables in pairs and notice that this leaves one table by itself. Groupings of three leave two tables adrift. Groupings of four again leave a lone solitary dinner party, and so on. My search for an even regularity in the arrangement of tables proves fruitless, and finally I surrender, forced to enumerate them. Seventeen tables, making seventeen parties of four, and thus, sixty-eight diners in all. Seventeen. A curious number. No matter how I try to organize the tables into a pattern of similar patterns, into clusters of equinumer ous clusters, I am foiled. One more or one less, and the reduction is possible. Sixteen tables could yield four groupings of four tables, or two groupings of eight tables. Eighteen tables would give three groupings of six, or two groupings of nine. Thus sixteen and eighteen are a composition of other, smaller numbers: composite, we might say. Seventeen, however, is not a number of numbers, it is not a pat tern of patterns. It is a first-order pattern, a primal pattern, a prime pattern. Primes and composites. It appears that the Greeks were the first to be aware of this distinction of the natural numbers. Perhaps it was a thoughtful shepherd clustering his sheep, or an acquisitive merchant counting out bolts of cloth, sheaves of wheat, or jugs of wine, who first discovered this distinction. The mystical mathematical musings of the Pythagoreans included a taxonomy of number dependent upon the concept of divisibility, whereby a given number can be obtained as some multiple of a specific number. Twelve is divisible by three, but not by five. The Pythagoreans called a number perfect, abundant (in the sense of having many divisors), or dificient according to its prop erty of being (respectively) equal to, less than, or greater than the sum of its divisors. For example, twenty-eight is perfect because it is the sum of its divisors: one, two, four, seven, and fourteen. The number twelve falls short of the sum of one, two, three, four, and six, and is thus said to be abundant, and eight, which exceeds the sum of one, two, and four, is deficient. Amicable numbers are two numbers defined by the property that the sum of divisors of the one is equal to the other. An example of such a pair is 220 and 284. Although the Pythagoreans were intrigued by the concept of divisibility, the first written record of classification into primes and composites dates from
The God-Given Natural Numbers
13
about 200 years after the death of Pythagoras and is found in the mother of all mathematical textbooks, Euclid's Elements. As with Pythagoras, the exact origins of Euclid (c. 330-275 B.C.E.) remain a mystery, but it is presumed that he was a student of Plato, or at least attended Plato's Academy, a scholarly retreat whose gate way bore the famous inscription "Let no man ignorant of geometry enter here." He then seems to have been called to Alexandria to teach mathematics at "the Museum," an institution of higher learn ing set up by Ptolemy I. Euclid was the author of several texts, but he is surely most famous for the Elements, which comprises thirteen books and is today thought of as primarily a textbook in geometry. It follows a largely axiomatic, or synthetic, approach to geometry in which "self-evident" geometric truths are laid down as axioms from which all of Euclidean geometry is derived through the application of Aristotlean logic. It is striking how much of Elements could today still be read and understood by any attentive schoolchild, and indeed, this axiomatic approach continues to be taught in many schools. Elements takes a geometric approach to number, whose study is mainly concentrated in Books 7 through 9. The Pythagoreans believed that number is all, but for Euclid and his followers, geometry ruled the universe and only those mathematical ideas that had geo metric realizations in terms of the perfection of ideal straight lines and circles (constructions capable of being realized with only the most basic tools of the working geometer: pencil, straightedge, and com pass) were of interest or even acknowledged. In such a mathematics, all investigations begin with a unit length that serves as the basic measuring stick against which all other lengths are compared or measured. These lengths are the geometric realiza tion of numbers. Addition of whole numbers corresponds to joining known lengths end to end. Multiplication of whole numbers is then achieved by repeated addition. It soon becomes evident that not all lengths (whole numbers) can be obtained by multiplying together two smaller numbers: that is, not all numbers are divisible. In this geometric setting, divisibility is expressed as the possibility that one number might be "measured" in terms of another (i.e. , that a given length could be "measured" as some number of copies of a given
14
STALK I N G T H E RIEMANN H YPOT H E S I S
length) . Thus, one number divides another if the corresponding length of the one "measures" the other. Among all lengths (numbers) those that are measured by no others are distinguished as the primes. PRIMES: THE INTEGRAL ATOMS
So here we have the fundamental dichotomy within the natural num bers: either a natural number is measured by another or it is not; either it is composite or it is prime. Herein lies the atomic nature of the primes, for if a natural number is composite, then we split it, rec ognizing it as the product of two necessarily smaller natural numbers. We continue the splitting, ultimately arriving at those prime (indeed, primal) pieces incapable of any further division. To give one simple example, the number 2,695 is split into the product 49 x 55, the former of which can in turn be split into the product 7 x 7, while the latter is the product 5 x 1 1 . The numbers 5, 7, and 1 1 cannot be split any further. Thus, given any natural number, by an analogous successive itera tion of splittings, continued so long as is possible, a collection of prime numbers is revealed, which if recombined by multiplication will yield the original number. In our example, 2,695 is the product of the primes 5 and 1 1 , and two copies of the prime 7. This is called its
primefactorization. Just as any chemical compound is composed of integral amounts of its constituent elements (exactly two atoms of hydrogen and one of oxygen make water; one sodium atom and one chlorine atom com bine to give us a molecule of table salt), so any number is produced as a product of particular primes, thereby yielding an arithmetic mole cule. However, whereas it is possible for two chemically distinct mol ecules to have the same molecular formula (such substances are called molecular isomers), a number's prime factorization is unique. This property of unique factorization is so basic to numbers that it is often called the fundamental theorem ofarithmetic. To uncover a fundamental element, the chemist takes a sample and subjects it to all sorts of tests: dabbing it with acid, holding it to a flame, boiling it in water. Physicists go a step further, looking for con-
The God-Given Natural Numbers
15
stituents that are even more fundamental, by bombarding materials with high-energy particles. What then of mathematics and the search for the primes? How to find these numbers that are measured by no others? How to mine the mountains of number for the irreducible, prime nuggets? It turns out that we need nothing as dramatic as a supercollider. All we need is a simple sieve, the sieve of the Greek mathematician and philosopher Eratosthenes. IN SEARCH OF P RIMES
Eratosthenes (c. 276-194 B.C.E.) was a native of Cyrene. Like Euclid, he was called to Alexandria, where he ran the library at the university and also served as tutor of the son of Ptolemy III. Eratosthenes is well known for his early estimate of the circumference of the earth, assum ing it to be a perfect sphere. Using only basic geometry and observations of the position of the sun, Eratosthenes arrived at a mea surement of about 24,660 miles, a pretty close approximation to the true value (a bit more than 24,90 1 miles). The "sieve of Eratosthenes" provides an orderly procedure for culling the primes from the natural numbers. The indivisible nature of the primes implies that once a prime is identified, none of its multi ples can be a prime. Since two is a prime, we can remove from consid eration all of its multiples-that is, all the even numbers-effectively "sieving" them from the body of natural numbers subject to our search for primes. We then look for our next prime among the first number not removed as a multiple of two, and it is three. Once again, we sift the remaining numbers for multiples of three, none of which are possible primes. Excluding the primes two and three, we now are left with only numbers that are multiples of neither two nor three, the smallest of which must again be prime-and 10 and behold, it is: the number five. We continue in this fashion, repeatedly sifting the integers with sieves of a size that grows as the primes grow. Implicit in Eratosthenes' sieve is one of the first primality tests: in order to determine whether a given number is prime, see if at any point in the sifting process it is removed from consideration. For the smaller natural numbers this test is useful; but for larger ones, those on the order of hundreds of
16
STALK I N G T H E RIEMANN H YPOT H E S I S
digits, such a simple-minded test is impossible to conduct. Poten tially, even the fastest computers of our day would require more time than the age of the universe to work on such a problem. Eratosthenes was simply interested in beginning to map the primes, and took a first stab at identifYing the primal archipelago dot ting the integer sea. Today, our interest in the primes is at least as intense, although perhaps more prosaic, for the primes provide the alphabet as well as the language that both enables and secures digital communication. SPEAKING IN P RIMES
In any single computer-or over the vast, highly distributed, inter connected network of computers that forms the Internet-resides a self-contained, logical world, whose language and phenomena find a common expression in number. This is the digital world, a silent world of a forever flowing stream of zeros and ones, strings of elec tronic fingers and toes, either bent at the knuckle or extended. These binary sequences are both its form and its function, providing instruction and carrying information. Zeros and ones make up the alphabet for the language used to describe, as well as to direct, events, providing a means of expression for the multiline program poems of executive orders. They are responsible for communicating our ideas, thoughts, conversations, and business transactions. The rules for manipulating the bits (binary digits) which are these zeros and ones are based on the smallest of prime cycles, the cycle of length two. Reasoning is distilled to a logic which becomes arith metic, capable of translation into complicated circuitry etched onto silicate wafers which serve to mediate the communion of thought and machine. These rules embody the distillation of debate and dialectic. In this stripped-down logic by which our machines live, we see mir rored the basic rhythms of life: breathe in, breathe out; yes, no; true, false; on, off. They are the distillation of us, simplified and pristine. Although the simplest of primes gives voice to the Internet revolu tion, it is the larger primes that enable an intelligible and credible dis cussion. The streams of information that course through the wires and airwaves are processed in huge chunks, whose sizes are deter mined by their prime factors. In this way the prime decomposition of
The God-Given Natural Numbers
17
a number is crucial to the efficient communication and analysis of the images, music, financial information, etc., which reside in our home computers and are sent across the Internet. Prime numbers are also fundamental to many of the processes that enable the reliable (i.e. , error-free) transmission of digitized informa tion. These are primes used in the service of error correction. The sim plest form of error correction appends a single bit to a binary message, setting it equal to zero if the message contains an even number of ones, and equal to one if the message contains an odd number of ones. That is, a single bit records the remainder obtained upon divid ing the total number of ones in the binary message by the prime two. That remainder is called the parity of the number of ones. The mes sage is now sent to the receiver, along with the parity bit. If during transmission a single bit of the message is changed, then the parity of the received message will differ from the received parity bit, indicating the message has been corrupted. More elaborate error-correction schemes use arithmetic properties of other primes. The biggest use of primes is found in the technology of digital cryp tography. This is the body of techniques used to ensure that informa tion is sent securely over the Internet. In this application, primes of tremendous magnitudes, on the order of hundreds of digits long, are used to encrypt messages in such a way that no present (or even fore seeable) technology would have a chance of deciphering them should they be intercepted. These cryptographic schemes start by interpreting the digital mes sages as numbers. One way in which this is possible is through the interpretation admitted by any finite list (string) of zeros and ones as a number's binary expansion. Reading right to left, any binary string represents an accumulation of successive powers of two in which a power is included only if the symbol in the corresponding position is a one. For example, the string 1 00 1 1 would be interpreted as the sum of one zeroth power of two (which is one), one first power of two, no second or third powers of two, and one fourth power of two. In short:
which is the sum of sixteen, two, and one, and gives nineteen. This is akin to our familiar decimal expansion, in which numbers read right
18
STALK I N G T H E RIEMANN H YPOT H E S I S
to left are shorthand for an accumulation of powers of ten. For exam ple, 2 1 9 is shorthand for nine zeroth powers of ten added to one first power of ten and two second powers of ten. Once a message is interpreted as a number, it can be scrambled by applying any of a variety of mathematical operations. The procedures that ensure the security of most Internet credit card transactions first repeatedly square a number (i.e., computing some very high power of the message-bearing number) and then divide this very high power by a previously specified number and record the remainder. This is done for various powers, and the final encryption of the original message is the number obtained by then multiplying together these numbers and taking one last remainder. The number of times this basic step occurs, as well as the number by which we divide, depends on some previously specified pair of huge prime numbers. In this way, the primes underlie all of modern digital technology, as the means by which information is made palatable to the computer. They are like the air we breathe-necessary, but superficially invisible, the ghost in the digital machine. Central to the cryptographic uses of primes is the underlying assumption that they are sufficiently plentiful. Quantification of their plenitude is precisely what the Riemann hypothesis is all about. So, as a first step toward its statement, let's start by trying to count the pnmes. EUCLID'S PROOF OF THE INF INITUDE OF P RIMES
We know that there is no largest natural number. Posit such a quan tity, and this conjectured supremum is defeated by the number obtained by adding one. That's what we mean by infinity. Like two children shouting in turn, "I dare you!" "I double-dare you!" "I triple-dare you!" the count could go on and on forever. But what of the numerosity of the primes? The totality of natural numbers is infinite, but do the primes need go on and on as well? After all, four simple molecules suf fice to create the seemingly infinite diversity of life cast within DNA.
The God-Given Natural Numbers
19
Astrophysicists believe that a finite number of basic particles are responsible for our presumably unbounded universe. Could it be that the primes too are finite? Euclid was the first person to demonstrate that the answer to this question is no. Euclid proved to the world that the primes are infinite by a clever argument which we might recast as a Socratic dialogue:
Euclid: Dan, I would like to impress upon you that, should we travel down the road of the number line, we might always find a prime number ahead. Dan: Indeed, wise Euclid, could this be so? How to be sure that no matter howfar we travel there is always a prime number ahead? Euclid: My dear Dan, as I travel along the number line I acquire prime numbers. At any point, I might stop to rest and wonder if indeed I now have them all. I might care to create a new number from thefinite collection in hand by multiplying together all ofthe primes in my possession, and then increasing this already large number by one. Dan: Yes, Euclid, you certainly could create that number ifyou like. Euclid: Now Dan, this large number is, no matter how large, still a number, and as such it must have an elementalfactorization in terms ofprime numbers. Isn't that so? Dan: Yes, that is most certainly so. Euclid: Do we agree that any prime number which is part of the factorization ofmy large number must divide that number exactly, so that there is no remainder? Dan: Absolutely, for that is what it means to be a factor. Euclid: Aha.' Here lies the problem. Should I divide my newly con structed number by any ofthe primes in my possession, we see that
20
STALK I N G T H E RIEMANN H YPOT H E S I S
we get a remainder not ofzero, but ofone! So, it cannot be that any ofthe primes currently in my possession divides our new number. Dan: Yes, that must be true. Euclid: Thus, there must be a new prime number out there on the number line, either that number out in the distance that J have just constructed, or another that lies somewhere in between here and there. Dan: Yes, that must be true. Let's illustrate Euclid's argument with an example. Suppose we stop our little stroll along the number line at the number thirteen, having identified the primes up to that point: 2, 3, 5, 7, 1 1, and l3. Euclid sug gests that we take a look at the number (2 x 3 x 5 x 7 x 1 1 x l 3) + 1 . A quick calculation reveals that the result is the number 30,03 1 . The number 30,03 1 cannot be divisible by any of the primes 2, 3, 5, 7, 1 1 , or l 3, because any one of these gives a remainder of one when we attempt to perform the division. For example if we try to divide 7 into (2 x 3 x 5 x 7 x 1 1 x l 3) + 1 we see that it goes into this num ber (2 x 3 x 5 x 1 1 x l 3) times, with a remainder of 1 . Nevertheless, as any natural number is either itself prime, or composed of a product of smaller prime factors, it must then be that the number 30,03 1 , while not necessarily prime, is at least made up o fprimes other than 2, 3, 5, 7, 1 1 , and l 3 . Euclid has discovered a procedure that takes i n a collection of prime numbers and outputs a new number which is composed of prime numbers that are not among our original collection. Through a sort of alchemy of arithmetic, the prime numbers 2, 3, 5, 7, 1 1 , and l 3 are mixed together by multiplication, and the result is incremented by one in order to produce the number 30,03 1 . When this number is subjected to the fires of factorization, two new primes are revealed: we write 30,03 1 = 59 x 509, and note that each prime factor is greater than any of the original prime ingredients. In this way, old primes are ground by arithmetic mortar and pestle to create new primes. It is a form of mathematical magic, perpetually producing primes.
~
The Shape ofthe Primes
EUCLID'S I NVESTIGATIONS
uncovered the fact that there are an infinity of primes. But this simple description sweeps much under the rug. Legend has it that the Inuit have a virtual avalanche of words to describe snow; similarly, mathematicians' familiarity and fas cination with infinity have led to an appreciation of infinity as a highly textured and nuanced notion. To say simply that the primes are infinite is, to a mathematician, not the end of a discussion but a beginning. "How infinite is it?" is a sensible mathematical question, and the way in which mathematics tackles the infinite is among the most distinguishing features of this SCIence. Once again, we can use the analogy of the study of the universe. We assume it is infinite, but to what extent? Some say the universe is ever expanding; others say that expansion will eventually give way to a great compression, the Big Bang folding back into the Big Crunch. These and other astrophysical theories depend upon the way in which matter and energy are strewn about the cosmos-the light and the dark, that which we can see and that which we can't. Are matter and energy spread about in such a way that eventually the expansion will halt? Or is there so little of each, so widely scattered, that the universe will grow forever? Where do matter and energy lurk and from where do they arise? Questions of form and questions of extent confront us. Similarly, we ask questions about the way in which the primes are distributed among the natural numbers. Are they speeding away, ever farther apart, or are they clustered together in recognizable constella-
22
STALK I N G THE RIEMANN H YPOT H E S I S
tions? In what sense are the primes infinite? Implicitly, these are ques tions of both quantity and quality, or more precisely, of cardinality and asymptotics. We'll look at these ideas now. CARDINALITY: HOW MANY P RIMES ARE THERE ?
On the surface, the declaration that the primes are infinite simply means that a list of the primes would go on forever. A first prime of two, a second prime of three, a third prime of five, a fourth prime that is seven, and so on and so on. Both explicit and implicit is the corre spondence that we create in our listing of the primes-a correspon dence attaching to any prime number that natural number which indicates its position as we list the primes in increasing order. Any collection of objects (or numbers) that can be put in such a correspondence is said to be countable, or denumerable. The act of constructing such a correspondence is the act of counting, or deter mining "how many" things are in the collection. The answer, a num ber, is called the cardinality of the collection, and the number representing this amount is a cardinal number. For example, our familiar positive integers are cardinal numbers, albeit finite ones. Three is the count of one, two, three, an intrinsic listing that exhausts any threesome. However, as Euclid first showed, our listing of the primes is never-ending and ever-reaching. Each prime patiently awaits its turn, so that a first description of the count able infinity of the primes is thus represented by a cardinal number different from any finite number. This cardinal number is usually written as �o, read "aleph zero." The iconography mixes the beginning of the Hebrew alphabet with the beginning of number to create a symbol for the cardinality of any infinite but countable collection. It is "beyond finite," and thus called a transfinite number. The theory of transfinite numbers was laid out by the German mathematician Georg Cantor ( 1 845- 1 9 1 8). He was the first to inves tigate arithmetic in the transfinite realm, showing that the countable infinity represented by �o was the smallest member of a new and mys terious transfinite number system. Since both the natural numbers and the prime numbers are countably infinite in extent, we see that transfinite accounting allows for the part to be equinumerous with
The Shape of the Primes
23
the whole. Moreover, even the collection of all nonnegative fractions, represented by all pairs of natural numbers, and hence seemingly much larger than the natural numbers alone, is still countable, a fact demonstrated by a listing of the fractions according to the size of the numerator and denominator. * It is not until we consider collections such as the set of all possible numbers between zero and one that we begin to find transfinite numbers which exceed a countable infinity. This fact is demonstrated by the consideration of any putative listing of the decimal expansions of all such numbers. Such a list cannot include that number which differs from the first number in the list in the first decimal place, or from the second number in the second dec imal place, and so on and so on. ASYM PTOTICS: H OW DO THE P RIMES G ROW?
The characterization of the primes as possessing a countable infini tude gives a quantitative description of their infinity, but what of the quality? For example, how does the countable infinity of the primes differ from that of the odd numbers, which also can be listed one by one? The odd numbers go marching on, regularly spaced at every other positive integer, but the primes seem to have no such simple rhythm. Twenty-three is prime; the next prime does not appear until six steps later, at twenty-nine; but after that the next prime is at thirty-one, only two steps farther. The primes twenty-nine and thirty-one are an instance of twin primes, prime pairs that are but two steps apart. Another such pair are the primes 227 and 229. Twin primes have been found way out in the numerical stratosphere, with the current largest pair having more than 24,000 digits. That is to say, these two numbers would require more than ten pages of this book to write down, both are prime, and they differ only in the last place. Numerical experi mentation supports the widely held but still unproved belief that there are an infinity of twin primes among the natural numbers. Such 'Such a lisring would begin (omi[[ing reperirions) like [his: 011 , 1 / 1 , 1 /2, 2/1 , 1 /3, 2/3, 3/ 1 , 3/2, 1 /4, 3/4, 4/3, 4/ 1 , . . (No rice [ha[ 2/4 is omi[[ed because 1 /2 is already included; and 4/2 is omi[[ed because 2/1 is already included.) .
24
STALK I N G T H E RIEMANN H YPOT H E S I S
brief paroxysms of primality are but one reflection of the irregular appearance of the primes. The rhythm of the primes is more a jazzy syncopation than the steady beat which describes the odd numbers. Edgar Allan Poe once wrote that "there is no exquisite beauty without some strangeness in proportion." The uneven footfall is at the heart of the allure of the primes. Thus, while both the odd numbers and the primes are countably infinite collections of numbers, the way in which they are distributed among all natural numbers differs. So now we ask a new question: Is there a way to describe the rate at which the primes occur? Can we give a law that predicts the appearance of the primes? Imagine the number line as an east-west sidewalk bordering a high way that extends from horizon to horizon, an infinite path marked evenly by integers, signposts spaced one yard apart. Euclid tells us that as we walk step by step, forward along the number line, no matter how far we walk we are always certain to come upon another prime. But when? Is there a roadside billboard claiming "Next prime, two integers ahead" or "Next prime, 2,000 integers ahead"? Prospecting for primes, we march forth with a basket in hand, examining each roadside integer, leaving in place those beautiful numerical hybrids that are the composites, looking only to pick the primes-which, in the words of the number theorist Don Zagier, "grow wild like weeds." As we advance, our basket grows heavy with primes, but the farther we go, the less frequently we appear to add to our collection. This is predicted by Eratosthenes' sieve, for as we stroll along the number line, each prime we encounter foreshadows an infinity of places ahead that cannot be prime. We reach two and toss it into our basket, knowing that of the numbers ahead, every other now is ruled out: four, then six, then eight, then ten, and so on. As we march forward from two, the first nonmultiple of two we reach is just ahead at three. As three is not a multiple of any of the primes cur rently in our basket, three must be prime, so it is added to our collec tion. As before, we look forward and anticipate the successive encounters every three steps ahead of multiples of three, another evenly spaced infinity of composite numbers. Thus does each en counter with a prime initiate a perfect wave washing away the regu larly spaced composites ahead.
The Shape of the Primes
25
We keep a tally of the contents of our basket and track the accu mulation of the primes. One prime in the basket as we pass two, then two primes in the basket as we pass three. The count remains at two until we pass the number five, whereupon we now have three primes in the basket. The contents of the basket increase to four as we pass the number seven. Then it is not until we go another four steps and reach eleven that we'll find another prime and thereby bring our tally to five, indicative of the five prime numbers that are at most eleven. If we knew the rate at which the primes appear, then we could also deduce the rate at which the size of the primes grows. For example, if we are told that there are ten primes that are less than thirty, but eleven primes no greater than thirty-one, then we know that the eleventh prime is equal to thirty-one. The irregular appearance of the primes reflects the varying intervals of time over which the contents of our basket do not change. We sense a rate of accumulation that generally appears to be progressively slower and slower, although intermittently interrupted by seemingly random hiccups of surprisingly short increments. How to quantifY the pace of the primes? A simple listing of numbers gives one picture of the accumulation of primes but still comes up short as a means of revealing any struc ture in their growth. Such a list is like a diary that purports to record the progress of a newborn baby through only a daily measurement of height and weight; better would be a picture book, allowing us to see and compare the day-by-day changes. Similarly, even better than our list of numbers is a picture. It is better to replace our description of the primes with a picture that might help reveal the pattern of accumula tion. A picture is worth a thousand words, and a good one here may be worth a thousand primes. So we make a mathematician's picture. We make a graph. An east-west axis crossed by another axis running north-south cre ates the setting for a picture that will tell the tale of the growth of the pnmes. Our number line heads off to the east, notched with the regularly spaced whole numbers. Measurements in the northerly direction serve to mark the number of primes in hand as we pass each natural number. So as we move eastward, the irregular appearance of the
26
STALK I N G T H E RIEMANN H YPOT H E S I S
30 25 � E
i5.. 20
"0
'U E " z
.D
15 10 5
20
40
60
80
1 00
1 20
1 40
Number line
Figure 1. Punctuated accumulation ofprimes as we move along the number linefrom 1 to 137. Above any position on the east-west axis (also called thex axis), the graph will be at a height equal to the number ofprimes that have been encountered up to that point.
primes gives a running count that builds a hip-hop staircase over the floor, the east-west number line. The height of the staircase at any position on the number line measures the number of primes found up to that instant. Primeless intervals, such as the short path from thir teen up to (but not including) seventeen are reflected as a plateau between these points. Here we will find a step, four units long. It rests at a height of six units, which reflects the six primes less than seven teen. But upon coming to seventeen, we find a new prime, so that our number of primes increases to seven, and so we add a new step to our staircase. Racing up this Dr. Seuss-like staircase would prove to be a bit dif ficult. In a rush we might hope to climb two or even three steps at a time. Were we hurrying up the evenly spaced steps of your office building or the local football stadium, we would soon acquire a rhythm, an even stride by which we would bound to the top. But no such luck here. We start with a step oflength one, which we follow by another, only to come in turn to steps of lengths two, and two again, then four, and then two, these lengths reflecting primeless stretches. Soon we slow down, seemingly unable to develop a cadence that will anticipate the broadening primal deserts. But we should not lose heart. Perhaps we are too close to the prob-
The Shape of the Primes
27
lem, as if Mendel had attempted to deduce a theory of inheritance from two pea plants, or Mendeleev had tried to write down a table of elements from the knowledge of only carbon and oxygen, or Darwin had tried to formulate a theory of development from looking at two finches. We need to go on and step back, in order to see the forest for the trees. SEEING THE FOREST FOR THE TREES
Although the minute variations seem hard to predict, order does appear to emerge from chaos if we look at things on a larger scale. In taking a step back we enter the realm of asymptotics, the study of long-term or large-scale behavior, capable of revealing the patterns that materialize as we "go to infinity." Out here, on the horizon, is where asymptotics comes into play. Just as we might distance our selves from life's minutiae in order to gain clarity, we do the same when we use asymptotics, choosing the view from afar in the hope that from this vantage point the moment-by-moment vagaries will coalesce into a clear trend. Asymptotics is the crystal ball of mathematical analysis, predicting the shape of things to come. We study phenomena and discern the broad outlines of what lies ahead, seeing the future all at once. Armed with this foreknowledge, we can then take the time to delve into the particulars of the past, getting our hands messy with the details that are sure to give birth to a structure already anticipated. Asymptotics enables the mathematician to understand mathematical life forward, in order to live it backward. Like the birth of our universe, the count of primes begins in chaos, a first three minutes of erratic behavior in which the initial primal materials are revealed; but then, slowly and surely, structure seems to emerge from the primordial mess. We leave the detail of that uneven staircase of the early prime landscape and zoom out and away, hur riedly snatching primes as we speed to higher and higher numbers. The effect on our picture of the growth of the primes is one of organi zation, like moving from a view of New York City at 1 ,000 feet, say from the top of the Empire State Building, to a view from a jet plane cruising at an altitude of 30,000 feet. Buoyed by this initial success, we continue onward and outward.
28
STALK I N G THE RIEMANN H YPOT H E S I S
1 50 1 25 � e
'&. a
1 00
''" ...
..0
e ::l z
75 50 25
200
400
600
800
1,000
Number line
Figure 2. Accumulation ofprimes up to 1,000. The uneven accumulation is still apparent in the staircase-like graph.
From 30,000 feet we move to a height of tens of thousands of miles, from jet plane to space shuttle. Jagged shorelines converge to a broad, smooth, almost cartoonish outline dividing land and sea, and so too does our picture of the accumulation of the primes move from step ping-stones to a gentle curve. As we step back, the structureless becomes structured; the inscrutable now seems within our under standing. As in the distant view of a pointillist masterpiece, the details have begun to fuse into a coherent whole that we now seek to describe. The every-otherness of the odd or even numbers means that they accumulate in a linear fashion, so that the picture of their count is a perfectly straight line. However, this sort of regular occurrence is not true of the primes, even from the long view. Taking hold of Eratos thenes' sieve, we might observe that generally, the farther out we go, the longer are the intervals between primes, for the farther we go, the more multiples of earlier numbers we find. This is in contrast to an infinite listing of multiples, where the distance between successive multiples is the same, no matter how far out you look. So, in the case of the primes, the long view of their accumulation yields not a line but a subtle curve. In order to understand the asymp toties of the accumulation of the primes, we will need to bend the straight line mathematically so as to match the growth rate of the primes. In order to do this we need to find another pattern of growth,
29
The Shape of the Primes
SO,OOO
60 ,000 '" "
E
'§.. a
'-
tl E ::l Z
40,000
..c
20,000
500,000 Number line
Figure 3. Accumulation ofpl'imes up to 1 million. It woks pretty close to a stmight line . . . but not quite.
another endless list of numbers, and use this second list like a sequence of local directions. They will indicate that the line might be modified by subtracting a little here, or adding a little there, or multi plying by a little here, or dividing by a little there-small perturba tions that accumulate to an ever so subtle deflection. Thus, what is needed is a way in which we might slowly pull our line down from the straight and narrow; not so much that it ulti mately heads to earth, but rather in such a way that it still heads out toward infinity, but at a less than speedy pace, as though we were gradually adding sand to the basket of a star-seeking hot-air balloon, to slow but not halt its journey. So, not a line, but the most gentle of curves-that is the coarse shape of the universe of primes, but what exactly is it? What is the mathematical law that will describe this slow and less than steady growth? This is the question that would spark Riemann's interest, years after it first excited two great mathematicians of the late eigh teenth century: Adrien-Marie Legendre and Carl Friedrich Gauss.
~
Primal Cartographers
FEW THEOREMS spring full-grown from the mind of a mathematician. Most research is a process of working out example after example in the hope of finding a common thread. Sometimes these examples take the form of reams of data that must be examined carefully in order to uncover a pattern of behavior. This is certainly the case in number theory, and especially in the asymptotic study of the accumulation of primes. The initial guesses as to the shape of its growth came from looking at great compendiums of data gathered in the form of ever longer lists of prime numbers, in the hope that the coastline, which appeared jagged from up close, might finally appear to some lucky investigator as an obvious, recognizable, smoothly growmg curve. The first cartographers of the primes were the French mathemati cian Adrien-Marie Legendre (1752-1 833) and the German mathe matician Carl Freidrich Gauss (1777-1855) . Among other things, both Gauss and Legendre are known for the tools they invented to calculate the motion of heavenly bodies. Their discoveries make pos sible the prediction of the positions of comets, planets, asteroids, and the like from a relatively small number of observations. This penchant for and skill at data analysis was what they each brought to the prob lem of predicting the location of the primes in the most distant reaches of the integer universe. Each would state a conjecture about the growth of the primes Legendre publicly and Gauss privately. Asymptotically, the two con jectures would amount to the same thing, but as befits the precision of
Primal Carto graphers
31
Gauss, his estimate would hint at the possibility of a formula for an exact counting of the primes. Gauss made a bold and inspired guess as to the shape of the prime territory, tossing down an intellectual gaunt let that Riemann would pick up some eighty years later, when he related the shape of the primal curve to the mysterious zeta zeros. LEGEN D RE
The first recorded prediction of the asymptotic growth of the prime counting function is attribured to Legendre. Of noble birth and broad intellect, Legendre moved quickly through the academic ranks, coming to the attention of the mathematical world in 1 782 for his prize-winning research on ballistics, and the determination of "the curve described by cannonballs and bombs, taking into consideration the resistance of air." Eventually, Legendre turned his attention to the study of phenom ena of a different order of magnitude, and his inventions for doing so live on to this day. Working on a cosmic scale, astronomers still use the eponymous Legendre fUnctions to help sort out the fine details in the nearly constant microwave background radiation that bathes the universe. This is the everlasting echo of the Big Bang, and in the minute fluctuations of the temperature lie keys to many mysteries of the origin and composition of the universe. Legendre's greatest contribution to astronomy, and to applied sci ence at large, may have been the invention of the "method of least squares," a technique invented to predict the motion of comets from a number of observations. These ideas, which he first set down in an appendix to a paper of 1 806 concerning the motion of comets, provide an important general technique for finding the simplest geometric shape (such as a line, curve, or ellipse) that most closely fits a collection of data. These data might be something as ethereal as the observed positions of a heavenly body, or something as mundane as the tick-by tick listing of a stock price or a history of corporate earnings. The Legendre functions and the method of least squares both exemplifY the interests, talents, and tools that Legendre had at hand when he turned his attention to the problem of predicting the rate of accumulation of the primes. Like a financial analyst looking to fit an
32
STALK I N G T H E RIEMANN H YPOT H E S I S
earnings record, Legendre was looking to find the best curve that would fit his data on the accumulation of the primes, a curve that could predict the future growth of these numbers. How to craft this curve? Like any good scientist, Legendre surely was driven by an aesthetic of simplicity, so he would be looking for a rule using only the basic instruments in the mathematician's curve cutting kit. As a starting point, Legendre could see, as we saw in Chapter 3, that the primal curve is very nearly but not exactly a straight line. * Legendre observed that the primes are distributed in such a way that the graph of their accumulation bends away from lin ear perfection. On the primal curve, as the distance of any point from the vertical axis increases, its height above the horizontal increases, but at a slower and ever-changing rate. Legendre posited that he could effect this slight modulation by allowing the vertical displacement to equal the horizontal displacement divided by an amount that grows ever so slowly as we make our way along the number line. This is the character of something that exhibits logarithmic growth, achieved in this case by using the logarithm of the displacement, and we now turn to its explanation.
The Rhythm ofthe Logarithm Logarithm. The very word can send shivers down the spine of even the most knowledgeable, but it is hardly fearsome, for as a rate of growth, called logarithmic growth, it is among the slowest of the slow. It is best understood in terms of its speedy cousin, the better known exponential growth, which derives from the exponential. Exponential growth is exemplified by the spread of a rumor told to two people, who at the next instant each tell two more people, who in turn then each tell two more people, and so on. This would be the way a tall tale might spread, lockstep, through a close-knit community: one person knows, then two more people know, then four more, then eight more, and so on. That is, first is two raised to the zero power, then it is raised *In the Cartesian plane a line is straight only if for each point on the line, the ratio of its distance from the horizontal axis to its distance from the vertical axis is constant. This number is the slope of the line.
Primal Cartographers
33
to the first power, then to the second power, then to the third power, and so on. In this perfectly synchronized town, the number of people just informed of the news would be described by two raised to a num ber equal to the number of instants elapsed. * There is nothing special about the number two. At each step, any person just informed of the rumor could have relayed it to three new people, or four or five or 1 00. The number of people newly informed at each step by any indi vidual is the base of this exponential growth process; the time elapsed provides the exponent with which we can measure the spread of the rumor. Thus, if each person relates the rumor to five new people, then at the fourth instant the fourth power of five, or 625, new people have just learned the rumor. In exponential growth, the farther we look, the faster something, such as money or information, will accumulate. The logarithm is in effect the exponential in reverse, for to grow as the logarithm means that the farther out you go, the more slowly things change. The logarithm is ubiquitous in the quantification of perception. Its rate of growth is well-suited to the relative nature according to which change seems to be experienced. For example, consider the Richter scale that is used to measure the strength of earthquakes. The Richter scale is such that our perception of the difference between tremors rat ing one and two on the Richter scale is the same as our perception of the difference between five and six. Nevertheless, a tremor that rates two on the Richter scale is in fact ten times as strong as a tremor that is rated one; and a tremor rated three is ten times stronger than a tremor rated two. Similarly, a tremor rated six is ten times as strong as one rated five. This constant ratio (of ten) between successive ratings on the Richter scale reflects the fact that the Richter scale measures the logarithm of the strength of tremors-i.e., it grows logarithmi cally in terms of the strength. A further implication is that six on the Richter scale marks a tremor that is 1 0 5 times more powerful than a tremor rated one. This in turn implies that the absolute difference between tremors rated one and two is very different from the absolute difference between five and six. *In the standard mathematical notation, 1 = 2°, then 2 = 2 1 , then 4 = 22, 8 = 23, and so on. In general at step n in the spread of the rumor, we will have just told 2" people.
34
STALK I N G THE RIEMANN H YPOT H E S I S
The former is essentially the difference between the numbers 1 0 and 1 00 (i.e., 90), whereas the latter is the difference between 1 0,000 and 1 00,000 (i.e., 90,000). However, in both cases the relative difference, which is given by the ratio of the two numbers, is the same: ten. Thus, while the logarithm of these two numbers grows by only one in each case, this increase of one reflects a much smaller absolute change for the smaller numbers (90) than for the larger numbers (90,000). The bigger you are, the slower you grow. Notice that in exponential growth (with a base of two), it took only a single time step to double the number of people just told a rumor from four to eight. By contrast, to move from four to eight on the Richter scale requires a tremor that is 1 0,000 ( 1 04) times stronger. The decibel system also works logarithmically. Our experience of a sonic increase of two decibels feels the same on the quiet end of the volume scale as at the loud end: turning up the volume on your radio from one to two feels the same sonically as turning it from ten to eleven. However, the absolute changes (in terms of the power of the sound wave crashing against our ears) are dramatically different. Quantitatively, the ratios of the power exerted in the change from one to two and in the change from ten to eleven are the same although the numerical differences are very different. To put it briefly, logarithmic scales quantifY the belief that perceived change is proportional to actual relative change, an idea that sometimes goes by the name
�bers law. Just as there are many possible exponentials, each depending on a given base, there are many possible logarithms. The standard choice of the mathematician is commonly called the natural logarithm. In accordance with mathematical tradition, we will simply call it "the logarithm." The natural logarithm undoes the exponential growth of the most commonly occurring base, denoted c. Although e may look extraordinarily uncommon, it is ubiquitous in mathematical and physical formulas. It is a number between two and three, more precisely at just above two and seven-tenths. * Thus it is not a natural number, like one, two, or three; but it is surely of nature. It is crucial to the formula that describes the shape of a loosely •A little more precisely, the decimal expansion of e starts our as e = 2.71 828 . . . It has most recently appeared in approximate form as the amoulU of money, in billions of
Primal Cartographers
35
hanging vine or necklace. In one of its most famous incarnations e is shown to be equal to the return on a dollar after one year, bestowed by an infinitely precise bank that continuously compounds interest at an annual rate of 1 00 percent. As the exponential races off to infinity, achieving galactic distances in very short times, it must then be that the logarithm, in an effort to undo this quick work, must quickly cut down to size even very large numbers. The twentieth power of e is nearly 500 million, so that the logarithm of a number near 500 million is close to twenty, and num bers near 500 million times 500 million still do not have logarithms much greater than forty.
A Near-Miss for Immortality Sidestepping the Prime Number Theorem On the basis of his recordings of the primes, Legendre proposed a for mula for counting the number of primes under a given value. We know there are four primes between 1 and 1 0, and eight primes between 1 and 20, but how many primes are there between 1 and 1 million? Legendre's calculations must have shown him that, roughly, the accumulation of the primes could be explained by a relatively sim ple statement, which would come to be known as the Prime Number
Theorem: The number of primes less than a given value is asymptotically that value, divided by its logarithm. Using this simple formula as a guess, we would estimate the num ber of primes less than 1 million to be about 72,000, and the number less than 1 billion to be about 48 million. Moreover, just as exact knowledge of the growth of the way in which the primes accumulate implies exact knowledge of the way in which the primes grow, so a coarse knowledge of the former yields a coarse knowledge of the latter. The implication of our proposed sucdollars, that the company Google hoped to raise with its IPO. For a wonderful his tory-and there is much to know-see Eli Maar's e; The Story ofa Number (Princeton Universiry Press, 1 994).
36
STALK I N G T H E RIEMANN H YPOT H E S I S
350000 300000 tl E 0-
"0 � " ..0
E " Z
250000 200000 -
Truth
' "
� " '''' ' "
,,-
"
1 50000 : Prime
1 00000 :
Number Theorem
50000
Number line
Figure 4. Gmph of the exact count of the primes ("T/'uth',), and the gmph that measures the ratio ofeach number to its logarithm.
cinct law of the accumulation of primes is that out on the horizon, an estimate of the size of, say, the millionth prime number is roughly 1 million times the logarithm of 1 million, or around 1 3,800,000. In other words, asymptotically, the size of the primes grows like the rank of the prime (i.e., where it is the list of the primes in increasing order) times the logarithm of the rank. * Figure 4 gives a head-to-head comparison of distant views of the graphs of the exact count of the primes and the estimate due to the Prime Number Theorem. While the former does underestimate the "truth," the amount by which it misses grows so slowly as to appear negligible at farther and farther distances. In order to quantifY this we can consider the ratio of the true count to the Prime Number Theorem's estimate. To say that these two are asymptotically the same is equivalent to saying that asymptotically, their ratio is close to one. Figure 5 bears out this fact. Had Legendre stopped at the simple formula of the Prime Number Theorem, he would have henceforth been known as the first person to conjecture (but not prove!) the Prime Number Theorem, and this
*In standard mathematical language, we would write the Prime Number Theorem as follows: The number of primes less than N is approximately equal to Nllog(N). Also: The Nth prime is approximately equal to N X log(N).
37
Primal Carto graphers
B
� E E ·c '= ti � E o u
�
� E
il t:S
1
______ __ __ ____ __ __ __ __ __ __ __ __ __ __ __ __ __ _ __ __
g "t;
u ..o " t: "
E
'O z
.g E c:2 J: E
Number line
Figure 5. Ratio ofthe true count ofthe primes to the estimate given by the Prime Number Theorem. It is almost exactly one.
seminal contribution to our understanding of the primes would have more than ensured his place in the mathematical pantheon. Instead, though, he viewed this rule as a rough template for the true law mea suring the count of the primes. So, like the financial analyst who, believing he sees a linear trend in earnings data, seeks to "fit a line" to those data, Legendre decided to search among a related collection of formulas for the best of the bunch. "Best" would be measured in terms of the ability of this graph to fit the data on primes that he had in hand, and so Legendre could use the same techniques that he had already developed to predict the future path of a comet streaking across the sky, now applying them to predict the course of the infin ity-seeking path of the count of the primes. Thus, in 1 808, in the second edition of his text Theorie des nombres, Legendre published an estimate for a count of the accumulation of the primes that was close to the logarithmically inspired ratio above.* 'To be precise, Legendre assumed that the rule describing the way in which the num ber of primes grows would be given by a law stating that the number of primes less than a given value Nis NI[A X 10g(N) + B] . He then found the values ofA and B that best fit the data he had gathered on primes. He concluded that this best fit occurred for A equal to 1 , and B equal to - 1 .08366, so that his recorded estimate for a formula pre dicting the number of primes less than a given number N was NI[log(N) - 1 .08366] .
38
STALK I N G T H E RIEMANN H YPOT H E S I S
Ironically, at the astronomical magnitudes where asymptotics rule, Legendre's small tweak of this simple ratio is not important, as its con tribution becomes increasingly negligible. Legendre had lost sight of the forest for the trees and made a classic error in data analysis by "overfitting" the data-finding a curve that fits well the data in hand while losing sight of the possible effects of the data that are still to come. This is a form of mathematical Monday-morning quarterback ing, in which perfect knowledge of the past too strongly influences a plan for the future. This brief instance of scientific shortsightedness cost Legendre one of the great mathematical discoveries of all time. Indeed, it was as though he had been on the verge of finding the North Pole but had settled instead on a different spot, just a few miles away. Legendre died in poverty, first losing his fortune in the aftermath of the French Revolution, and later losing his government pension after coming out on the wrong side of a political battle. The last several years of his life seem to have been rife with fights over principle; in particular, he spent a great deal of energy battling for recognition of the priority for his research achievements, among which was his guess at the growth rate of the primes. Once again he would come up short, trumped by the same mathematician with whom he had quarreled over priority for the discovery of the method of least squares: Carl Friederich Gauss. GAUS S : PRINCE AM ONG MEN
Gauss is judged by many to have been the greatest mathematician of all time. Although he showed significant mathematical aptitude at a young age, philology and theology were his primary intellectual inter ests until he was well into late adolescence. It was only at the urging of his teachers that he finally decided to focus on mathematics. Gauss's reputation comes from his contributions to the full diver sity of modern mathematics, both pure and applied. Nevertheless, his nickname, "prince of mathematicians," speaks to a particular facility for the "queen of mathematics"-number theory. So it is perhaps an oedipal embarrassment that Gauss is considered the father of modern number theory. His book Disquisitiones arithmeticae was the primary
Primal Carto graphers
39
reference in the field for decades after his death and served to intro duce many a mathematician to the subject. But number theory is just the tip of the Gaussian iceberg. Even today, there is scarcely an area of mathematics that does not continue to feel his influence. He is often credited as being the first mathematician to give analytic work (i.e., the calculus) a rigorous formulation. His achievements in geometry include the discovery of non-Euclidean geometries, as well as the invention of modern differentialgeometry, a subject which, at Gauss's prodding, Riemann would later remake. All this is but a brief nod to Gauss's purer work, and the problems and techniques of applied mathematics were also close to his heart. In particular, astronomy was both an inspiration for and a primary ben eficiary of his work in applied mathematics. Like Legendre, Gauss was keenly interested in the problem of predicting the trajectories ofheav enly bodies from finite lists of astronomical observations. Gauss remains famous for predicting the orbit and consequent rediscovery of Ceres, a relatively small asteroid that had disappeared from sight. He took it upon himself to predict its position from its known observations and from Newton's law of gravitation. This achievement brought the twenty-four-year-old Gauss worldwide sci entific acclaim in the sciences. To facilitate the pages of necessary cal culations, Gauss developed a shortcut that was the forerunner of an idea now known as the fast Fourier transform, or FFT. The FFT is one of the most important computational tools ever invented, and one of the key concepts enabling modern digital technology. The other tool of which Gauss made a great deal of use was the method of least squares, and while he made no effort at all to publicize this grandfa ther of the FFT, he did take great pains to claim the method of least squares as his own, thereby initiating the long fight with Legendre for priority.
Gauss and the Primes Gauss's failure to announce his invention of the fast Fourier trans form was fairly typical of him. He is as famous for the results he didn't publish as for those he did. Gauss's strict personal criteria for publication are summed up in his often quoted remark, "One does
40
STALK I N G T H E RIEMANN H YPOT H E S I S
not leave the scaffolding around the cathedral.» This was his way of saying that he would permit the public to see a result only when he himself understood it so well that he could support it with a self contained, beautifully structured proof, free of jury-rigged reasoning and flimsy argument. With such rigid requirements for publication, much of what he accomplished was relegated to his Notebooks, a trea sure trove of research results that would have served as a publishable life's work for almost any other mathematician. Gauss's work on the distribution of the primes must have seemed to him to be swathed in scaffolding, for the only evidence of his research on this subject comes from a letter written to a former stu dent, the German astronomer Johann Encke (1791-1865). Encke was a well-known scientist in his own right: his name lives on in the Encke Gap, a region found among the rings of Saturn; and in Encke's Comet (first discovered by the astronomer Jean-Louis Pons in 1 8 1 8), whose return Encke accurately predicted along with the periodicity of its orbit. Encke was the director of the Berlin Observatory, and in this capac ity he undoubtedly had a good deal of contact with the astronomi cally-interested Gauss. Evidently, at one point Gauss and Encke had discussed the prob lem of the distribution of the primes. In the famous letter, dated Christmas Eve, 1 849, Gauss appears to be responding to a conjecture by Encke on the asymptotic growth rate that counts the primes. It is a safe bet that Encke had in his possession a more extensive table of primes than Legendre had, and that from these he found a new esti mate of the count, one which varied slightly from Legendre's overfit conjecture.* Gauss may still have been smarting from his fight with Legendre over the priority of the discovery of the method of least squares, or perhaps he was motivated by his penchant for perfection. In any case, he claimed in his letter to Encke that fifty-eight years earlier (which would have been almost twenty years before the publication of Legendre's estimate) , he had already estimated the rate of growth of *Encke used N/ [(log N) 1 . 1 5 1 3] ( 0 estimate the number of primes less than N, as opposed (0 Legendre's estimate of N/[(log N) 1 .08366] . -
-
Primal Cartographers
41
the accumulation of the primes, and he believed his estimate to be much better than that of either Encke or Legendre. In this letter Gauss said that as a boy of fourteen or fifteen he had taken great pleasure in counting the primes, culling them from inter vals of thousands (called chiliads) . He wrote, ''As a boy I had consid ered the problem of how many primes there are up to a given point." From his computations he had determined that the "density of primes" near any given number "is about the reciprocal of its loga rithm." The notion of density used here by Gauss is akin to our under standing of population densities. In demography, population demity refers to the average number of people living in a given unit of area (e.g., a square mile) or a particular region (e.g., a state) . A density of this sort will usually vary as the region varies, in either size or location. For example, the population density of South Dakota is quite differ ent from that of New Jersey. Similarly, we can ask how many prime numbers "live in the neigh borhood" of a particular number. Gauss's estimates imply that as we traipse along the number line with basket in hand, picking up primes, we will eventually acquire them at a rate approaching the reciprocal of the logarithm of the position that we've just passed. For example, at around 1 million (a number whose logarithm is about thirteen) we find primes at a rate of about one in every thirteen numbers; at 1 bil lion, we've slowed down to something like a rate of one prime for every twenty-one integers; and so on. What we now want to do is find a way to convert the ever-slowing rate at which the primes appear into a count for the number of primes that we've acquired up to any given point. Notice that the rate at which primes occur gets slower and slower the farther out you go, but the rate at which this slowing occurs is itself very, very slow-so slow that it hardly seems to change from interval to interval. In other words, over very big intervals, the rate at which we find primes doesn't change much. This is what we'll use to try to estimate the number of primes up to a given amount. For example, the numbers between 990,000 and 1 , 0 1 0,000 all have a logarithm dose to thirteen, so on average, about one out of every thirteen numbers between 990,000 and 1 , 0 1 0,000 are prime. Thus the
42
STALK I N G THE RIEMANN H YPOT H E S I S
exact number of primes in this interval would be about one thirteenth of the interval size, or one-thirteenth of 20,000, which is a little over 1 ,000. Another way of looking at this is that for the num bers between 990,000 and 1 , 0 1 0,000, each number "contains" about one-thirteenth, or approximately the reciprocal of its logarithm, worth of a prime. Extrapolating this argument, in order to get the number of primes between one and some given point, you might as well simply add all these reciprocals of the logarithms for all the num bers from one to the chosen stopping point. Gauss knew how to get a very close estimate of this sum using cal culus, for the idea described above is the same idea that goes into determining distance traveled solely from knowledge of the rate at which you are accumulating distance, i.e., the velocity. This is the technique of integration. By using these tools he was able to compute an estimate of the prime-counting function given by the logarithmic integral, mathematical shorthand for adding up these tinier and tinier logarithmic reciprocals. Using the logarithmic integral as an approximation to the count of the primes, Gauss determined that for primes up to 3 million his esti mate was in general better than Encke's, but worse than Legendre's. Nevertheless, given the trend in the differences, Gauss expected that eventually (i.e., asymptotically) his estimate would be much closer. Gauss was confident in his estimate, but because it had only empirical justification, he relegated it to his private personal papers. Figure 6 bears out his confidence, showing that although Legendre's estimate is more accurate initially, Gauss's is more consistent overall, and eventu ally is better.
The Shape ofThings to Come Gauss's formula does indeed predict that in the long run, the number of primes less than a given value is that value divided by its ratio: i.e., the exact statement of what would come to be known as the Prime Number Theorem. This is also true of Legendre's guess. However, out on the asymptotic horizon, relatively small differences are ignored. This means that two estimates can be asymptotically the same, but along the way the amount by which they differ from the true count
43
Primal Cartographers
x
Legendre
Encke
Gauss
1 ,000,000
45
464
1 28
2,000,000
42
797
1 20
96
1 , 1 62
3,000,000
----- ------
1 53 --
4,000,000
1 77
1,541
205
5,000,000
130
1 ,782
1 24
6,000,000
27 1
2,204
226
7,000,000
263
2,471
1 77
8,000,000
351
2,829
22 1
9,000,000
361
3, 1 05
1 86
1 0,000,000
560
3,566
338
Figure 6. Amounts by which Legendl'e's, Encke's, and Gauss's estimates ofthe numbel' ofprimes less than a given amount differftom the exact count. The successive I'OWS show the differencesfor the numbel' ofpl'imes less than 1 mil lion thl'ough the number ofpl'imes less than 10 million. For example, the sixth row indicates that Legendl'e's estimateJor the numbel' ofprimes less than 6 million differs ftom the exact count by 271; Encke's estimate differs by 2,204; and Gauss's estimate (using the logarithmic integl'l1l) is best, diffel'ing by only 226.
can be very different. Gauss's estimate would prove to be much closer to the true count than Legendre's estimate. The points on the upper curve in Figure 7 chart the growth of the difference between Legendre's estimate of the number of primes up to a point and the true count, and the lower curve does the same for Gauss's estimate. For example, Legendre's estimate of the number of primes less than eighty million (8 x 1 07) is off by about five thousand, while Gauss's estimate is off by about five hundred, and generally, Gauss's estimate is much better. Notice that Legendre's error looks to be speeding off to infinity, while Gauss's seems to stay relatively con stant. In truth, it too will eventually get arbitrarily large, albeit at a much, much slower rate. Moreover, Gauss's exact formula for estimating the primes, this log arithmic integral, holds much more information about the growth of the count of the primes than the rough guess given by the asymptotic.
44
STALK I N G T H E RIEMANN H YPOT H E S I S
6000 :
Gauss
Number line
Figure 7. Cornparisons ofthe amounts by which Legend/'e's estimate differs fi'Om the exact count ofthep/'imes and the amount by which Gauss's estimate differsfrom the actual count.
Legendre's guess shrouds the shape of the primes in a formless sack cloth, while Gauss's prediction is more a silk robe and gives the asymptotic arc a Christo-like covering that not only outlines the gen eral behavior of the primes in the large but also gives some hint as to the true variation lying beneath. Legendre's estimate is a dead end; Gauss's is a beginning, for embedded in Gauss's estimate of the number of primes by the loga rithmic integral are the first hints of a refinement of the coarse esti mate that is the Prime Number Theorem. Figure 8 takes a closer look at the difference between the true count of the primes and Gauss's estimate out to 1 0 million. If in keeping with the asymptotic spirit of things we neglect the fre netic local ups and downs, a clear slowly growing curving trend is vis ible. This represents the first detail in the description of the growth of the count of the primes, the first tantalizing stirrings of the lively vari ation missed by the Prime Number Theorem. The shape underlying this first variation must itself be determined if we are to uncover the true count of the primes. Indeed, after obtaining this next level of truth, there is still more to do. A next best approximation begets the
45
Primal Cartographers 7000 -c
;J " a ·0 � .'" § " Cl '" OJ 1:
OJ
1: OJ
6000 '" u a .;:
0..
� 'O i:: " w
0 u _
" .: u '" "
.LO OJ u
1:
tIE 6
5000 4000 3000 2000 : 1 000 :
Number line
Figure 8. Absolute difference between the true count ofthe number ofprimes less than a given amount and Gauss's estimate. Riemann would set his sights on distilling the implicit underlying curved trend.
need for another and then another. Getting to the truth at the end of this process is what would excite Riemann and would be one of his great achievements, but he wouldn't do it by himself. Leonhard Euler and Lejeune Dirichlet would provide the first indications of the right direction of investigation.
~
Shoulders to Stand Upon
MATHEMATICS , like any intellectual discipline, proceeds by a process of accumulation, the discoveries and assertions of one generation building on those of its predecessors. The great scientist and mathematician Sir Isaac Newton once famously said, "If I have seen farther than others, it is because I have stood on the shoulders of giants." Such is the process of science, the discoveries of one genera tion serving as a legacy for the next. Some scientists leave tangible, hard-won treasure chests of data; others might leave a map or a cryp tic deathbed remark, hinting at a land of untold promise and reward. Many the budding scientist hears the story of the apple falling on Newton's head, jarring a connection of ideas in which Newton saw the apple's analogy in the moon moving around the earth, or the earth around the sun, each forever falling toward yet never colliding with some massive attractor. This scientific legend is a romantic version of what was surely a hard-fought intellectual assault that placed Newton atop a pyramid of astronomers, mathematicians, and physicists. For he succeeded in developing a theory whose foundations reach back to the ancient scientist-priests who first laid out the heavens according to constellations and then organized their motions in the first space-time theory: the calendar. Fast-forward to the sixteenth century, when a law of elliptical orbits discerned by Johannes Kepler from the careful and copious astronomical observations ofTycho Brahe would serve as inspiration but would ultimately be recast by Newton as a conse quence of a general law of gravitation, whose derivation would require the additional invention of a new mathematical science, calculus.
Shoulders to Stand Upon
47
Newton's is but one story in the saga of science, and it is a story which is replayed in our tale of the primes. Newton's laws predicted the motion of the planets; Riemann's law will predict the appearance of the primes. Newton had his Kepler and Brahe; Riemann had Legendre and Gauss. Newton's calculus can trace its roots as far back as the Greeks but builds directly upon that of his friend, mentor, and predecessor at Cambridge, John Barrow. In order to move beyond Gauss, Riemann would push off from the Greek-inspired work of the great Leonhard Euler, as well as that of his own friend, mentor, and academic predecessor, Gustav Dirichlet. Together they would show the way, bringing the tools of calculus to bear on the problem of the primes. They would provide the first sextants for the primal explorers. EULER: A B RI D GE-BUILDER
Leonhard Euler ( 1 707-1 783) was a Swiss mathematician and one of the all-time greats. Prolific in every sense, over his lifetime he pro duced thirteen children as well as forty-eight volumes of work, much of which is still being edited and mined for hidden gems. For Euler, like Gauss, and then Riemann, the various areas of mathematics were all of one piece-a startling range of achievement in light of what is today a highly Balkanized world of science. There is scarcely an area of mathematics that was not touched by Euler's intel lect. The great mathematician Pierre-Simon Laplace is reputed to have said that the only way to learn mathematics is to read Euler, "the master of us all," a title by which Euler is still known. Euler moved across the mathematical landscape, creating whole disciplines when necessary, building bridges between what had appeared to be isolated islands of research. The mathematics used today to model the natural world is rife with the name Euler. Scien tists still study the Euler equations in order to understand the com plexities of hydrodynamic flow, and models of the evolution of climate and phenomena like El Nino make use of Euler-Galerkin
methods. A mathematical Milton, Euler lived his last eighteen years in dark ness, blinded by cataracts, but continuing his ceaseless work through dictation to his son, Jacob. During these last years of his life, Euler
48
STALK I N G THE RIEMANN H YPOT H E S I S
laid to rest one of the then great unsolved problems in celestial mechanics, the mathematical subject which studies and models the motion of the heavenly bodies. Euler was able to calculate the orbit of the moon as influenced by the gravitational pull of both the sun and the earth. This was the first instance of a solution to Newton's three body problem (the two-body problem is effectively solved by New ton's equation for the gravitational attraction of two masses upon each other) . Though blind, Euler was still able to see in his mind's eye the movement of these heavenly bodies and to make that movement visi ble to the rest of us in a complex mathematical poetry of equations. Awake beneath an internal night sky, Euler spun an epic tale of the moon held in thrall to the forces of gravity and so forever falling toward the earth and sun, while Milton, in his own blindness a cen tury earlier, turned his thoughts to God and told us of the fall of man, simultaneously pulled toward paradise and hell. With these sorts of achievements Euler succeeded in finding simple but relevant models for real-world phenomena. Relevant, but not exact, for rarely is any real part of the world exactly modeled by math ematics. One purpose of abstraction is to create a model that extracts the essence of a phenomenon while ignoring the details of the partic ular instantiation. Perhaps no area of mathematics more aptly epito mizes this outlook than topology, a subject whose creation is often credited to Euler. As Euler defined it, topology-or, as it was classically called, geome tria situs ("the geometry of position")-is concerned with those prop erties of position that do not take magnitude into account. Geometry requires the notions of angle and distance, or a metric, and in this sense is rigid. Objects are differentiated by microscopic changes. Topology is a more forgiving subject. Objects that can be deformed continuously one into another are considered one and the same. A bagel is a doughnut is a wedding band is a hula hoop. But a bagel is not a pretzel. The beginnings of topology are most often traced to Euler's solu tion of a famous little puzzle, the seven bridges of Konigsberg. Euler determined that it was impossible to find a walk through the city's sys tem of bridges (part of its parks system) which would traverse each bridge exactly once. His solution of this brainteaser reveals the
Shoulders to Stand Upon
49
breadth and depth of his intellect. He strips away the real-world trap pings to focus on what is important. Landmasses are shrunk to ideal points, or vertices, and bridges are shrunk to lines, or edges. The beau tiful parks system of Konigsberg is replaced by a graph. This graph is different from the graph of a function: it is, instead, a representation which retains only the relevant relationships of connectivity. Euler's investigations mark the beginning not only of topology but also ofgraph theory, a subject which has been revivified in the modern analysis of networks. The Internet is a huge graph; wires become edges; hubs are vertices. We send e-mail to one another, flooding landlines and airwaves with a rush of bits and bytes. Fittingly, the man who laid the mathematical foundations for the study of the conduits of this electronic tsunami is also the one who first modeled the simple wash of water from one end of a pipe to the other. These great achievements have in common an inspiration drawn from the world around us, but Euler also drew inspiration from within, and he was just as excited to investigate the intrinsic nature of mathematical objects. This is the man who wrote the book Anarysis of the Infinite, and his other nickname, "master of the infinite," speaks to the delight with which he pursued the study of the infinite series, a cornerstone of the calculus. Moreover, it is through the infinite series that Euler connected number theory with analysis. In so doing, he provided Riemann with a first major clue for attacking the puzzle of the distribution of the primes.
Connecting the Dots: The Continuous Meets the Discrete While mathematics does share characteristics with its scientific cousins, there are also differences. Certainly one way in which mathe matics differs from the physical sciences is that mathematics produces theorems, not theories. A theory explains that which we can currently see, and makes predictions of things yet to be seen, events whose appearance will at best confirm an idea, and at worst prove it wrong. Aristotle's description of the basic constituents of matter as fire, earth, air, and water eventually falls to Lavoisier's notion of elements. This in turn gives way to the atomic theory of Dalton, which steps aside for Bohr's electron-nucleon model, which ultimately yields to Gell-
50
STALK I N G THE RIEMANN H YPOT H E S I S
Mann's world of quarks of various strangeness and flavors. A theory of the heavenly motions is similarly evolutionary: Ptolemy gives way to Copernicus, who steps aside for Newton, who tips his hat to Einstein. On the other hand, 3,000 years ago Euclid proved that there are an infinite number of primes, and his proof stands up today and will stand up tomorrow and forevermore. It is a fact not of a time, but of all time. Theorems are stuff for the ages, but a theory is only as good as its most recent experimental confirmation. The eternal nature of mathematical truth is irresistible to some, a sliver of immortality which in its cartoonish extreme has grown into the portrayal of mathematicians as arrogant pedants in films like A Beautifol Mind and Good Will Hunting, or plays like Tom Stoppard's Arcadia and David Auburn's Proof Conversely, thoughts of eternity can also be daunting. The proof rendered almost always seems so messy in comparison with the pure beauty of the perfect theorem; as a result, stories of Gauss-like perfectionists abound in mathematics. Perhaps this is the source of the aesthetics of brevity in mathematical proof The shorter the argument, the closer it seems to the economy of thought that is the flash of insight, which is intuition. Nevertheless, for all its trappings of eternity, a proof of a mathe matical fact does not necessarily lay to rest a subject or problem. We may know something to be true, but it is not necessarily then set aside like a specimen in a museum, hidden away only to be admired or trot ted out as needed. The best theorems mark not the end of a quest but the beginning, and perhaps the frequency with which a known fact is revisited is the best measure of its importance or centrality. The pre fect portraits of Vermeer do not make unnecessary the canvases of Picasso or Rothko; the sensuous figures of Rodin do not obviate the forms of Brancusi or the wraithlike bronzes of Giacometti. Each great artist makes his or her own comment on the human form or the human experience in a different way, revealing a different aspect of the truth of existence. Powerful, groundbreaking work is less the actual accomplishment than the future work that it enables. It is less the sight than making possible the ability to see anew. Proofs are themselves mathematical objects, and so a proof, ifbeau tiful, if useful, will show the way to many more interesting results, suggesting connections between areas and new techniques to solve
Shoulders to Stand Upon
51
different problems. In many ways, mathematics is less the body of theorems than the reasoning by which they are supported. It is more means than end, more trail than mountaintop. Euclid revealed the infinitude of the primes, but his approach lacked subtlety. Plowing straight ahead like a bulldozer, he had effec tively bypassed the more delicate questions of the nature of this infini tude. Given any finite collection of primes, Euclid had shown that there was a way in which a prime larger than any of those could be uncovered. While this implies infinity, there is no part of the explicit procedure which is itself innately infinite. Two larger questions How infinite are the primes? How fast do they grow?-are fundamen tally asymptotic questions, i.e., questions about the long-term behavior of infinite processes, and questions that the infinity-phobic Greeks would not ask. In this way Euclid's proof is certainly of his time. It wasn't until 2,000 years later, in Alpine Switzerland, that Euler found a new route to this mountaintop of number theory through the subject of analysis. Analysis, like the calculus from which it derives, is the study of the continuous and the infinitely small. Euler's great con tribution to our story of the primes is as a bridge-builder, for it was Euler who first saw a way to connect the mathematics of the calcu lus-smooth and continuous-to that of the primes, which is jagged and discrete. He had the imagination to reverse Euclid's telescope in order to see the primes anew.
An Upside-Down Look at the Primes Euler turned Euclid's proof of the infinitude of primes on its head. Instead of looking directly at the prime numbers, Euler considered their reciprocals-one-half, one-third, one-fifth, one-seventh, one eleventh-and asked a different question: What happens if I start to add these fractions? Euler realized that if he could understand how this sum accumu lated, it might be possible to begin to understand the rate at which the primes appeared. For example, suppose that the primes were very, very rare, so that as we traveled far, far out along the number line, the vistas between the primes generally became increasingly enormous.
52
STALK I N G T H E RIEMANN H YPOT H E S I S
This extreme sparsity would imply that in a listing of the primes, the individual numbers soon would be huge. These huge primes produce very tiny reciprocals. Primes of a size on the order of millions have rec iprocals on the order of millionths, and so on. So, if it were the case that primes were fairly rare, then a sum of prime reciprocals com puted by the successive acuumulation of reciprocals of the primes taken in increasing order soon would be growing by only infinitesimal increments. Conversely, were distances between primes not tremen dous, then the reciprocals would not be decreasing as rapidly, and the accumulation of their reciprocals would not tail off so quickly. In short, the rate at which the primes arise would have dramatic implica tions for the way in which the sum of their reciprocals grows. This sum of prime reciprocals is an example of an infinite series, and their study is part of the foundation of the calculus, the toolbox of mathematical techniques developed by Newton to help tame nature. Newton used this sort of infinitesimal point of view to under stand both the curves of nature and the motion of objects as a succes sion of better and better finite approximations, which in some ideal limit would realize the natural phenomena under investigation. The steady accumulation of number that is this infinite sum of prime reciprocals can once again be visualized as a walk along a number line. Each reciprocal of a prime may be interpreted as a directive to walk forward by some fixed amount. First walk half a mile, then another third of a mile, then another fifth of a mile, and so on. What can we make of such a journey? Maybe it is like the steady march described by orders to travel one mile, then another mile, and then another mile, and continuing this forever, a journey in which we tromp off to the horizon well aware that with patience we will one day reach, and then pass, any goal which we set for ourselves. On the other hand, perhaps it is instead like the seemingly infinite journey to the other side of the room mapped out as a directive to first walk half the way, then from this new position half the way again, and on and on, resulting in an infinite incremental journey that eventually comes as close as we like to the other side while never reaching the destina tion; our goal is the gold ring that is forever in sight, but forever out of reach. In this way the world of the infinite can be rife with frustra-
Shoulders to Stand Upon
53
tion, confusion, and paradox, which the ancient Greeks found so dis turbing that they forbade its study and discussion. Their "abhorrence of the infinite" is traced in part to thought experiments like this never-ending half and then half again trip across the room, which the philosopher Zeno used to construct paradoxes capable of "prov ing" the impossibility of motion through a misunderstanding of the infinite. Rather than a source of confusion, such considerations were a fount of pure joy for Euler. He delighted in working with the infinite, and often succeeded in making mathematics out of its mysteries. Zeno's half the way, then half again walk is an example of an infinite series that converges, and the corresponding sum of one-half to which is added one-fourth, and then one-eighth, and so on is given the sum (or "limit") of one. By contrast, the steady unbounded march is an example of a divergent series. Euler bushwhacked out on this mathematical frontier. Disregard ing the rules of polite mathematical behavior, he acted somewhat as a mathematical outlaw, often making intuitive leaps without sweating over the details that would consume his followers. Euler paid little attention to the rigor and cries for certainty that would be the hall mark of the great German mathematical analysts, and in this border land between truth and proof, he managed to stake out a new homeland for the world of mathematics. This would be a place in which number theory would flower, seeded by Euler's study of the distribution of the primes through the analysis of a particularly fas cinating infinite series, the harmonic series, which would ultimately provide the raw material with which Riemann fashioned the key to understanding the primes.
The Harmonic Series Euler's analysis of the infinite accumulation of prime reciprocals started with a mathematical fact that had been known to mathemati cians for hundreds of years: if a person tried to add up all the recipro cals of the integers, adding one-half to one, then one-third to this, and then one-fourth to that, and so on, then (just as in the case of an infi nite series modeled on the steady mile-up on-mile walk to the hori-
54
STALK I N G T H E RIEMANN H YPOT H E S I S
zon), this sum of reciprocals would ultimately though very slowly sur pass any preassigned amount. This summation of all reciprocals is called the harmonic series, honoring its relation to the Pythagoreans' numerical-musical mysti cism. The Pythagoreans discovered that two strings which are com mensurable in length (meaning that the lengths are simultaneously whole multiples of a single common length) will sound harmonious when struck simultaneously. For example, one string might be twice as long as the template, and another string three times as long. These lengths are commensurable and, in particular, are in a ratio of two to three. Continuing along this path of harmonious commensurability, we build a many-stringed guitar by snipping off as many jointly har monious lengths of string as we please: start with one string, then add another whose length is one-half that, then add another whose length is one-third the first, and so on. At some point we stop adding strings, and the chord we get by simultaneously strumming all those strings will be music to our ears. The harmonic series describes how much string would be needed if we wanted to build an imaginary guitar using strings of all possible reciprocal lengths. To see that indeed there isn't enough string in the world, consider the successive amounts of string needed for the first two strings, then the next two, and then the next four. Counting lengths off in feet, our first string will require one foot and our next string another half foot. The next two strings are one-third and one fourth of a foot respectively. Now we make a simple observation: since one-third is bigger than one-fourth, the addition of these two strings requires that we use at least an additional half-foot of string. We now quickly count off the amount required for the next four strings. It is exactly the sum of one-fifth, one-sixth, one-seventh, and one-eighth foot. Since each of these four lengths is greater than or equal to one-eighth, their total length is greater than or equal to that of four pieces of string each of which is an eighth of a foot long, which would be another half a foot. So, in order to make the next four strings we need more than another half-foot of string. At the next step we will look at the total length needed for the next eight strings; after that, the following sixteen strings; and so on, at each point doubling the number of strings we considered at the previ-
55
Shoulders to Stand Upon
ous step. If we keep going like this, we find that we continue to accu mulate string in chunks that exceed one half-foot of string. * This means that our Platonic guitar requires an infinite amount of string, or equivalently, that the harmonic series grows to infinity.
A Harmonious Prooffor Euclid Where do the primes enter into the harmonic series? How can it be that the chord played from the harmonic series has embedded in it a music of the primes? Well, just as all integers can be created by multi plying together all possible finite collections of primes, so too the har monic series, which is the orderly accumulation of the reciprocals of each and every integer, can be expressed not just as a sum but also as a product-a product requiring an infinity of factors, one for every prime. This is the famous Eulerfactorization of the harmonic series. Each factor comprises the entire contribution to the harmonic series corresponding to a single prime number, and is itself another infinite series that is for each prime much like Zeno's half and then half again walk. This walk of one foot, then half that, then half that again (for another quarter), and half that once more (for one-eighth), and so on, which comes as close as we like to advancing a whole two feet, makes sense of the statement that the infinite sum of the recipro cal powers of two is two: 1 + 112 + 114
+
118 + . . . = 2
Similarly, were we instead to walk one foot, then one-third that, then one-third that again (for one-ninth of a foot), and one-third that again (for one twenty-seventh of a foot), and so on, we would approach as close as we like a distance of one and one-half feet. For one last example, the same procedure using a factor of one-fifth at each step brings us ever and ever closer to a point five-fourths of a foot away. In short, we say that the sum of all inverse powers of three is one 'What we are edging (Oward here is that at step N of this process, the length of each of the next 2N strings is at least 1 I2N + I , so that their (0tal length is at least the product of these two numbers, or 1 12.
56
STALK I N G T H E RIEMANN H YPOT H E S I S
and one-half, and the sum of all inverse powers of five is one and one quarter. We do this for every prime, considering walks where at each step we advance a fraction of the distance that we advanced just one step before, with the fraction equal to the reciprocal of the prime. In each case (as for two, three, or five) there is a point which we approach as close as we like, but never surpass, and this is the sum (or infinite series) of the reciprocal prime powers. These are the factors of the Euler factorization, one piece for each prime, and when they are all multiplied together their product yields the song that is the harmonic senes. This is a song of infinity, for, as we've already seen, the harmonic series grows without bound. So were there but a finite number of primes, then the product of their terms (each ofwhich is finite) in the Euler factorization would necessarily be finite. For example, were five the biggest prime number, then, using Euler's idea, it would be possi ble to express the harmonic series as the product of the three numbers derived above: two (for the sum of the reciprocal powers of two), three-halves (for the sum of the reciprocal powers of three), and five fourths (for the sum of the reciprocal powers of five) . Their product is thirty eighths. This is a fact that is terribly at odds with the divergence to infinity of the harmonic series. Since mathematics cannot tolerate two contradictory facts, there must be an infinite number of prime terms. Thus, once again are we assured of the infinity of the primes. Euler has found a new road to a truth first uncovered by Euclid.
Building a Series ofPrimes Out ofLogs Perhaps this whispered hint of primal melody in the harmonic series would have been enough to lure mathematicians to investigate it, for a new proof of a known important fact is always of interest. But Euler went even further. In doing so, he showed his awareness of the deep connections between the rate at which the primes grow and the prop erties of the harmonic series and its generalizations, making a great leap in an intellectual and mathematical journey that would find its culmination in the work of Riemann. Euler rolled up his sleeves and dug into his new proof of the infini-
Shoulders to Stand Upon
57
tude of the primes. Using his mastery of the manipulations of the infi nite, he was able to use the Euler factorization to extract from the har monic series only those terms (or string lengths) of prime reciprocals, such as one-half, one-third, one-fifth, and one-seventh, thereby con sidering only that Platonic guitar whose strings have lengths corre sponding to prime reciprocals. The relative sparsity of primes as we travel along the outer reaches of the number line might lead us to believe that as we add only these reciprocals (adding one-half to one-third, and then one-fifth, one seventh . . . ) the numbers we are adding get small so quickly that, as in the case ofZeno's reluctant room-crosser, these sums all remain for ever before a fixed bound. But no. Instead Euler discovered that like the harmonic series, the sum of the reciprocals of the primes also passes any preassigned point. But the rate at which we get to these signposts, or accumulate our guitar strings, is so slow as to require the patience of a yogi. The harmonic series is already slow enough, growing roughly at the rate of the natural logarithm: the sum of its first ten terms is equal to roughly the natural logarithm of ten, which is about three; and the sum of its first 1 million terms is roughly equal to eighteen. But even slower than slow, a plan of accumulation described by adding prime reciprocals grows (and we use the word lightly) at a rate that is like the logarithm of the logarithm. Even after 1 million terms, we have accu mulated only the logarithm of the logarithm of 1 million, or the loga rithm of eighteen, which is a little bit more than four! This is a growth rate whose progress is like that of the snail's snail, or the slug's slug, a rate that is slower than slow.
One Series among Many Through his use of the harmonic series, Euler built the first bridge that would connect the discrete world of number to the continuous world of calculus. This is a first analytic understanding of the way in which the primes are distributed. The growth of the accumulation of their reciprocals must tell us something about the rate at which they appear at the far reaches of the number line. But Euler's contribution goes beyond being the first one to put a
58
STALK I N G T H E RIEMANN H YPOT H E S I S
stopwatch on the primes. Euler, Master of the Infinite, saw the har monic series not as an isolated piece of infinite amusement, but rather as one possibility among a vast and never-ending continuum of infi nite series. He started by considering the sum of the reciprocals of the natural numbers, which is the harmonic series, but why not other related accumulations too? What about the squares of the recipro cals? The cubes? Why not consider the sum of the reciprocals of any power?* It turns out that the behavior of the prime reciprocals is different from all these other sums. For example, suppose that only the recipro cals of square numbers are added, a walk across the room for Zeno's persistent student that demands first a step of one foot, followed by one-fourth (i.e., one over two squared) of a foot, and then another ninth, and so on, each time moving ahead by an amount equal to the reciprocal of the next squared number. In fact, this sort of walk does stay within the realm of the finite, never surpassing (and coming as close as we like) to a total distance of one-sixth 1( (pi) squared, a number that is slightly larger than one and one-half. Similarly (and mysteriously), walks that use other even pow ers (such as directing that at each step you move forward according to the reciprocals of fourth powers, or sixth powers) yield other fractions of powers of 1(. The implication is quite surprising: since the sum of the reciprocals of the primes passes any point, but the sum of the reciprocals of any fixed power is finite, then it must be the case that the primes are more densely distributed than any power. This is one of the first recorded facts regarding the density of the primes. By considering the sum of the reciprocals of the squares, and the cubes, and so on, Euler showed that in fact the harmonic series was but one species in a newly discovered genus of infinite series, that the 'In other words, Euler looked at series like the sum of reciprocal squares,
the sum of reciprocal cubes,
and more generally, the sum of reciprocals raised to any natural-number power.
Shoulders to Stand Upon
59
harmonic series was but one point of an infinite family of infinite series, one for each natural number. Riemann would have first seen this reading Euler's Analysis of the Infinite as a schoolboy. It would ultimately prove to be the starting point of his own reinvention of the study of primes, but it was Riemann's true academic mentor, Gustav Dirichlet ( 1 805-1859) who would be the first to show that Euler's bridge between analysis and number theory was more than a footpath. DIRICHLET: FI RST ONE OVER THE BRIDGE . . .
By all accounts Dirichlet was an amiable man, a child of the middle class, the son of an educated woman and the local postmaster. A pre cocious youngster, he graduated from secondary school at the age of sixteen and left Germany to study mathematics in France, then the center of mathematical research. Dirichlet had the repeated good for tune of finding one generous mentor after another. He made steady progress in his work and was soon able to return home, where his successes in France eased his progress up the codified stages of the German academic ladder. He assumed a highly coveted professor ship in Berlin in 1 8 3 1 , the same year in which he married Rebecca Mendelssohn, a sister of the composer Felix Mendelssohn. Dirichlet was able to revitalize what had become a rather moribund mathematical environment in Berlin, his intelligence and affability serving as a magnet for the mathematical community at large. After Gauss died, Dirichlet assumed the professorship of mathematics at Gottingen and remained there until his own death.
Periodic Tablesfor the Primes Dirichlet enters our story as the first one to pick up on Euler's use of the infinite series in the study of primes. Euler saw the harmonic series as one of a family of possible infinite series, one for each natural num ber, corresponding to a consideration of the sums of the reciprocals of the squares, cubes, fourth powers, and so on. Dirichlet took this pow erful idea one great leap further, realizing that this discrete collection of infinite series built from the harmonic series could be filled out to a
60
STALK I N G T H E RIEMANN H YPOT H E S I S
continuum's worth of series. He did this in order to lay to rest another of the giant open problems of number theory, the problem of finding "primes in arithmetic progressions." He was thus able to begin to clas sify and understand the diversity of the primes. An arithmetic progression is a list of integers that we arrive at through walks along the number line of regular step-size. The simple walk of step-size one starting at one yields a list of all the positive inte gers, and thus a list in which an infinity of primes occur (since all of them occur there). Should we start at two and then bound along by steps of length two, we find ourselves jumping from even number to even number, thereby obtaining the entire list of even numbers as an arithmetic progression of step-size two, starting at two. If we started at one, a step-size of two would take us from odd number to odd num ber, thereby producing the odd numbers as an arithmetic progression. With respect to the primes we notice a dramatic distinction between these two arithmetic progressions. Outside the starting point of two, the list of evens contains no primes, whereas the list of odds contains all primes other than two, and in particular contains an infi nite number of primes. Now we try a new game, instead choosing to hop along the integers taking three steps at a time. If we start at one, we move on to four, then seven, and so on. Starting at two we leap to five, then eight, and so on; and finally, starting at three we go to six, then nine, and so forth. The last of these arithmetic progressions gives all the multiples of three: 1, 4, 7, 1 0, . . . 2, 5, 8, 1 1, . . . 3, 6, 9, 12, . . . Euclid had asked if the list of primes went on forever, i.e., if among the list of natural numbers there were primes as large as we would desire. We can ask the same question of our arithmetic progressions: in any arithmetic progression, are there primes as large as we desire? To begin, we notice that of our three progressions with step-size three, the last, which comprises only multiples of three, can contain no primes other than the prime three. More generally, we see that we
Shoulders to Stand Upon
61
must restrict our attention to those progressions in which the step-size and the starting point have no factors in common. Euler was the first to wonder publicly about the possibility of an infinity of primes in an arithmetic progression (outside the caveat mentioned above). He believed it to be possible but never proved that it was. Legendre claimed to have proved it on two separate occasions, but both times his proof turned out to be incomplete. It was Dirichlet who finally showed that in such regularly spaced romps through the integers an infinity of primes would be encoun tered. In order to do this Dirichlet looked to Euler. Dirichlet wanted to mimic the way in which Euler had proved that there were an infin ity of primes. Euler had accomplished this through the use of the Euler factorization, showing that the infinite nature of the harmonic series necessitated that it be made of an infinite number of factors, one for each prime. Dirichlet wanted to do the same, but this time to make an infinite series whose factorization used only primes that occurred in a given arithmetic progression. The collection of infinite series dreamed up by Dirichlet (consisting of one series for each arith metic progression) to arrive at his results has come to be known as the Dirichlet L-series. Whereas in his other work on the primes Euler had only seen fit to use natural number exponents in his variations on the harmonic series, Dirichlet saw that he could study the arithmetic pro gressions by filling in the spaces on the number line, allowing himself the possibility of using sums of integer reciprocals raised to a power that could be any real number greater than one. * With this he was able to prove that an infinite number of primes appear in any arith metic progression, except for those in which the step-size and starting place share a common factor. POSTSCRIPT
Euler's new proof of the infinity of the primes and Dirichlet's further exploitation of Euler's techniques to find deeper structure within the *In other words, Dirichlet allowed himself to consider infinite series that looked like 1 + 1/25 + 1/Y + 1/45 + 1/55 + . . . where s could be any real number greater than one.
62
STALK I N G T H E RIEMANN H YPOT H E S I S
prime landscape revealed the power of analytic techniques in under standing the primes. Along the way, a torch illuminating the primes passed from one occupant of the Gottingen professorship in mathe matics to another. The next handoff would prove to be the decisive one. Riemann was waiting.
~
Riemann and His "Very Likely" Hypothesis
IT IS 1 859, a landmark year in the history of science. Charles Darwin publishes The Origin ofSpecies and thereby publicly announces a scientific theory capable of characterizing the growth and development of the natural world. His theory of evolution arrives on the scene as an intellectual thunderbolt, the culmination of a steady accretion of data and data analysis, using the notions of fitness and selection to explain the minute variations as well as grand temporal trends recorded in this naturalist's detailed notebooks. This same year, a bit east of the other side of the English Channel, the earthshaking achievement of a diffident British biologist finds its mathematical counterpart in a brief paper, "On the Number of Primes Less Than a Given Magnitude," presented to the Berlin Acad emy by a German mathematician, as shy and softspoken as Darwin. The accompanying lecture, delivered in honor of his election to the rank of "corresponding member" of this august body, marks the reve lation of a new idea for explicating the growth patterns implicit in the tables of primes and composites, the flora and fauna of the Platonic world of the natural numbers. This "Darwin of number theory" surprises the leading scientists of the day by announcing the discovery of a beautiful formula, one that can illuminate the irregular hiccups which mark the punctuated evolution of the primes and can also illuminate their fusion into a broad overarching picture summarized by the conjectured, but still unproved, "Prime Number Theorem." The discoverer of this formula is Bernhard Riemann, and his paper contains what is the now famous
Riemann hypothesis.
64
STALK I N G T H E RIEMANN H YPOT H E S I S
Like its Darwinian doppelganger, Riemann's paper is an epochal scientific achievement. With an outlook akin to that of his teacher Gauss, Riemann sought a description of the distribution of the primes capable of more than a coarse-scale replication of their accumulation. But like any aspiring scientist, Riemann surely wanted to do more than chisel one more detail into the cathedral of the past achieve ments of others. His goal was no less than an illumination of the whole truth, from the large to the small, from the far-off asymptotic grand sweeping curve to the zigzag nooks and crannies of close detail. His discovery of an exact formula, with the concomitant implications of his hypothesis is, like Darwin's theory of evolution, one of the great scientific signposts, a place from which we measure a before and an after in the history of ideas. THE ROAD TO THE RIEMANN HYPOTHESIS
Unfortunately, we have no written record detailing the thought process that guided Riemann during his work on primes. We have in hand some of his notes, as well as his paper, but while not every math ematician suffers Gauss's need for cathedral-like perfection, most mathematics papers are largely "scaffolding-free," often stripped of any signs of intuition or thought processes. Striving more for clarity and leanness of argument than narrative, they are mostly assertion and proof and often frustratingly free of motivation. Riemann's diary (ifhe had kept one) might have tracked the day by day and week by week wanderings of his intellectual journey, provid ing a road map of his frustrations and reasoning, his disappointments and happy surprises. Instead, although we have the fruits and some signs of the labor, we have relatively little insight into the process. We admire the masterful, inspired sculpture; we feel the heft of the chisels and mallets; but the muse eludes us. There is often an element ofluck, a serendipitous collision of the right problem and the right problem solver. Surely there is a lot of hard work, for that is the way things usu ally go in mathematics: perspiration almost always trumps inspira tion. In a search for a deeper understanding of Riemann's work, the number theorist Carl Ludwig Siegel (1 886-1 981) made it his business to wade through the reams of paper containing Riemann's posthu mous notes (the famous Nachlass, or estate) . Legend has it that upon
Riemann and His "Very Likely" Hypothesis
65
receiving one package of notes and sifting through page upon page of scribblings and scratchings, overwritten formulas and calculations, crossings-out and erasures, Siegel turned to the mathematician G. H. Hardy and joked, "So here is Riemann's great insight." Ultimately, what we are left with is the product of a quiet unconscious formation of a web of surprising connections among raw materials of scholarship acquired over a lifetime. The collection of Riemann's mathematical output is almost as terse as his manner-and his life, which was cut short by tuberculosis when he was just shy of forty. His published writings fill only a slender vol ume, two-thirds of which appeared posthumously. He lived to see only nine of his papers in print, of which just five were devoted to pure mathematics. Nevertheless, the ratio of quality of work to quan tity is extraordinary. The progress of modern mathematics and physics was forever altered by these relatively few papers, among which was the jewel that led to his reinvention of the study of the dis tribution of the prime numbers. This would be a true but terse math ematical tour de force, displaying an economy of thought well suited to a man of boundless imagination but with no time to waste. In its brevity and unfinished form, it is like a metaphor for Riemann's own unfinished life.
Beginnings Riemann's lecture at the Berlin Academy was one shining moment for him in a year replete with personal achievements. At the age of thirty one he had just been named professor of mathematics at the Univer sity of Gottingen, thus following in the footsteps of Dirichlet and Gauss. This pedigree reflects the professorship's status as one of the most prestigious of the few and highly sought-after senior mathemat ics posts in Germany. Riemann's appointment must have been as much a source of relief as of pride. For years he had been the sole financial support for a large extended family, a situation which had stretched thin his meager academic salary. Now Riemann would have the financial and professional security necessary to allow him to turn full attention to his work. Riemann was a man of humble beginnings, the second of six chil dren (two boys and four girls) born to a Lutheran minister and the
66
STALK I N G THE RIEMANN H YPOT H E S I S
educated daughter of a court councillor, then living in Germany's Hanover region. His early studies were directed toward following in his father's footsteps, but an innate intellectual affinity and talent for the sciences (as well as a shyness dramatically at odds with the duties of a minister) soon overcame filial devotion. With his father's approval, Riemann took up a formal study of mathematics and physics. Riemann cut his mathematical teeth on the classics. Legendre's The ory of Numbers, the massive two-volume text containing Legendre's asymptotic estimate of the accumulation of the primes, served as an introduction to number theory, and is most likely the original source of Riemann's interest in the distribution of the primes. Legend has it that he devoured Legendre's treatise in just one week, returning it to the surprised lender with a word of thanks and the offhand remark, "I know it by heart." Riemann's own work on the distribution of the primes would knit this adolescent mathematical inspiration with the tools of the calcu lus, acquired from his other early mathematical influence, Euler's Introduction to Infinitesimal Analysis, one of the first calculus texts. Although this work was dated in approach by the time Riemann read it, the many of inspired and inspiring manipulations of beautiful, infi nite formulas (including the wondrous harmonic series and many of its relatives) would serve him well in his later investigations. Riemann's graduation from a Gymnasium (secondary school) was followed by enrollment at Gottingen University, where he hoped to study with Gauss. Unfortunately, the quiet Riemann and the arro gant, often acerbic Gauss must have been a poor match in tempera ment. After one year Riemann took leave to continue his studies in Berlin, at that time a friendlier and livelier place, recently reinvigo rated by the leadership of the avuncular Dirichlet. Two years later, more mature and confident, Riemann returned to Gottingen, and in 1 8 5 1 he finished his doctoral dissertation under Gauss and began his professional mathematical career.
A Reinvention ofSpace At the time of the lecture in Berlin, Riemann was perhaps best known for his work in geometry, the roots of which can be found in his paper
Riemann and His "Very Likely" Hypothesis
67
"On the Hypotheses That Lie at the Foundation of Geometry." Like "On the Number of Primes . . . " this is a brief work-only nine pages-but it bubbles with the primordial material of an entire intel lectual world. It is more sketch than finished work, more discourse than disquisition, written pithily in a tone and style appropriate to a man who perhaps feels that he has not the time to say it all precisely, and so chooses to say only the most important things in as rich a lan guage as possible. The genesis of "On the Hypotheses . . . " is one of those happy intellectual accidents that seem to dot the landscape of scientific achievement. Riemann wrote this paper in order to complete his "Habilitation," a piece of postdoctoral research required for advance ment through the torturous trellis that was, and still is, German acad eme. As was the custom, Riemann submitted to his supervising committee a list of three possible topics, the last of which was an investigation of the "foundations of geometry." Traditionally, an examinee would be asked to pursue his or her first suggestion. How ever, as fortune would have it, the chair of the committee was Gauss, whose interest in the possibilities of geometries other than Euclidean geometry (in secret, he had already discovered the existence of non Euclidean geometries) probably motivated his unconventional choice of this last topic for Riemann's research subject. Riemann took up the challenge and produced a paper containing all the basic ideas of what is now known as Riemannian geometry. The classic example of a Riemannian geometry is the geometry describing the surface of a perfect sphere. Its apparently Rat "local" neighborhoods (think of how the surface of a beach ball would appear to a gnat, or how the surface of the earth appears to a human being) are stitched together in such a way as to produce a domain in which even our familiar triangles acquire puzzling properties. The sphere's constant positive curvature implies a land in which the angles of a tri angle sum to an amount greater than the Euclidean or "Rat" paradigm of 1 80 degrees. For example, consider the three-sided region obtained by choosing two distinct points on the equator and traveling due north from each of them, thereby creating two lines that meet at the North Pole, and then closing the spherical triangle by joining the original equatorial points by a line hugging the equator. The lines due
68
STALK I N G T H E RIEMANN H YPOT H E S I S
north form angles of90 degrees with the equator, so that whatever the angle they form upon their intersection at the North Pole (somewhere between zero and 1 80 degrees), the measures of the angles of this tri angle sum to something greater than 1 80 degrees. Furthermore, since any point on the sphere's surface can be described using two numbers, or coordinates (e.g., latitude and longitude), this Riemannian manifold is two-dimensional A sphere is the simplest sort of Riemannian space, providing the prototype of a space of constant curvature, but because of this con stancy it is a geometry that hardly shows off the power of Riemann's invention. For Riemannian, not Euclidean, geometry is the measure of the world around us. The rigidity of Euclidean geometry is well suited for perfect lines, angles, and circles, but Riemannian geometry quantifies the local variation of the real world. Riemannian geometry makes sense of a topography that includes craggy peaks and smooth dales, rolling hills and flat deserts. But the true power of Riemannian geometry lies in its ability to describe worlds beyond sight, spaces that shimmer at the edge of our imagination. The great geometers seem to have access to a realm beyond the senses, but just within the reach of the intellect. These thinkers are Homeric in their invention ofa language able to illuminate worlds to which all must be blind. Riemannian geometry is one such great mathematical and artistic tool. By enlarging the notion of coordi natization from the two-dimensional east-west, north-south Cartesian plane to an arbitrary number of dimensions, the language and tools of Riemannian geometry mediate the familiar intuitions of two- and three-dimensional phenomena to guide investigations in spaces that are beyond direct sensory experience. This geometry brings Euclid down to earth, while simultaneously pushing Descartes to the stars. In particular, Riemannian geometry has proved able to describe the phenomena of a relativistic world and the geometry of the space-time continuum, thereby anticipating and enabling the discoveries of physicists like Einstein and Stephen Hawking. It gives mathematical life and precision to the infinite and infinitesimal, enabling a descrip tion of a curved universe of gravitational mountains and valleys, pockmarked with black holes and looped in knots of stringlike ten drils of energy.
Riemann and His "Very Likely" Hypothesis
69
Riemann's sole contribution to the theory of prime numbers is in essence a creation of a Riemannian geometry of the landscape of primes. He would find a way to express the count of the primes that provides the detail of their one-by-one accumulation without missing the overall growth, a way of showing forest and trees all at once. Thus Riemann pushed ahead of Gauss and laid the groundwork for a rigor ous justification of Gauss's conjectured Prime Number Theorem. The range of tools and techniques that Riemann would bring to bear on this problem make it seem as if he had been preparing all his life for this moment; his work on the primes seems to be a nexus for a lifetime of mathematical experience. In taking on the challenge of understanding the distribution of the primes, he would pick up where his thesis adviser Gauss had left off. Riemann reached back to the writings and conjectures of Legendre that had first piqued his adoles cent interest in the primes and made use of a mature application of the formulas and manipulations found in his first calculus book, writ ten by Euler. In so doing he would extend the work of his friend and mentor Dirichlet. These advances in the study of complex numbers, complex analysis, and Fourier analysis are crucial to the Riemann hypothesis, as well as to the modern study of the phenomena of sounds and signals. We now turn to them. A COM PLEX D I S SERTATION Riemann's doctoral dissertation marked the beginning of his career as a professional mathematician. Gauss was the chair of Riemann's doc toral committee, and Riemann's work was sufficiently important and novel to raise even the dour Gauss's eyebrow. In signing off on Rie mann's degree, Gauss remarked that his dissertation "offers convinc ing evidence . . . of a creative, active, truly mathematical mind, and of a gloriously fertile originality." With this work Riemann effectively announced himself as a founder, along with the French mathematician Augustin-Louis Cauchy ( 1 789-1 857), of complex analysis: the extension of the tech niques of calculus to the realm of complex numbers. This discipline is responsible for some of the most important mathematical tools in physics and engineering. Riemann's paradigm-shifting insight was the
70
STALK I N G T H E RIEMANN H YPOT H E S I S
realization that it is applicable to the study of the distribution of prime numbers.
Complex Numbers Complex numbers are, in some way, the final destination of a logical evolution of the notion of number. As we have already seen, the con cept of number has its origins in what Kronecker called the God given natural numbers. As descriptors of discrete sensory experience, the natural numbers are but a small intellectual step away from the consideration of their ratios. These ratios, numbers like one-half or two-thirds, make up the rational numbers, whose visual and tangible realization in terms of Euclid's commensurable lengths retains the familiarity and immediacy of their "natural" progenitors. Following these rational developments things begin to get more complicated-perhaps crazy, but irrational at the very least. The Greeks discovered that there are naturally occurring numbers which cannot be expressed as the ratio of two natural numbers. Make a square, and consider the ratio of the length of a diagonal to any of the (equal) sides. It turns out that this number, which is the square root of two, cannot be represented as the ratio of two natural numbers. We have found in nature our first irrational number. Initially, the Greeks must have been troubled by the existence of the irrationals. These numbers may be ever better approximated by ratios, yet are unrealizable as such, and so embody the notion of the infinite-thus they would have been anathema to a people with a declared abhorrence of infinity. However, the Greeks' displeasure was surely tempered by the possibility that some of these numbers might indeed be given geometric descriptions. Upon further investigation, some irrational numbers came to be viewed as "better" than others, in the sense of being more intuitively appealing. The ancient Greeks subscribed to a philosophy of the necessity of the tangible nature of number and therefore admitted only those numbers capable of con struction using the basic instruments of the geometer-straightedge and compass. Such constructible numbers include the square root of two, represented by the line drawn along the straightedge lying on a diagonal of a square whose sides have length one; they also include 1(,
71
Riemann and His "Very Likely" Hypothesis
realized with the help o f our trusty compass as the circumference o f a circle with diameter equal to one. Over time, this prejudice in favor of tangibility abated, and the explicit constructions embodied in geometry eventually gave way to implicit ones. A notion of number grew beyond that which is visually apparent to include that which we wish to be true. Soon, numbers were anticipated according to their ability to satisfy some mathemati cal property, a wish couched in an emotionless and universal language of equations, written as letters, numbers, and other much more exotic icons. The now familiar stalwarts of arithmetic-zero and the nega tive numbers-were at one time viewed as impossibilities because they lacked substance. But describe something, and it will come. For example, what of the number "minus one"? We envision a "quantity" which has the property that when added to two, it gives one, summa rized by the symbolic sentence x + 2 = 1 . Similarly, we dream of a number called "zero," defined by the property that when it is added to any number at all, it is without effect. No one would deny that these early gedanken numbers have turned out to be quite useful. We can continue the implicit constructions that yield numbers like zero and the negatives, and arrive at a notion of number that includes instantiations of even more fanciful dreams: numbers defined in terms of relations among their powers. What of the number whose second power is equal to two? Or equivalently, what of a number whose square when added to minus two yields zero?* That is the now familiar square root of two. What of a number whose third power when added to twice its second power and then further diminished by seven gives zero?t Numbers like these are related to, but more general than, square roots, cube roots, and the like, and have the property that multiples of their powers can be balanced in such a way as to produce zero; they are called algebraic numbers. Numbers that do not fall under this umbrella are said to be transcendental, as the impossibility of their implicit description using only a finite number of powers somehow puts them beyond the apprehension of mortals. Although most real 'I.e., a square root of two is the number that when substituted for x in the expression XI 2 produces O. In other words, it is defined as a number that satisfies the equation XI 2 = O. 'I.e., a number that satisfies the equation :x:3 + 2il 7 = O. -
-
-
72
STALK I N G THE RIEMANN H YPOT H E S I S
numbers are transcendental (there are but a countable infinity of alge braic numbers within the uncountably infinite number of reals, leav ing an uncountable number of transcendentals), it is notoriously difficult to establish the transcendence of any particular number. The proofs of the transcendence of e (by the Frenchman Charles Hermite in 1 873) and 1( (by the German Carl Lindemann in 1 882) are still hailed today as two of the great achievements of modern mathematics. Within the classification of numbers as algebraic or transcendental, still more surprises await us. Algebra has opened gates onto vistas that seem beyond our intuition. Among the algebraic numbers occurs one whose definition seems quite natural: What of that number whose square is minus one? What kind of number could this be? Why should such a number be considered any different, any more "com plex," than the analogously defined square root of two? In spite of their common algebraic origins, it is a square root of minus one that bears the name imaginary number. The real numbers and imaginary numbers together produce the collective of complex numbers. In An Imaginary Tale, Paul Nahin tells the wonderful story of the intellectual birth of the complex numbers. In fact, having a number whose square is equal to minus one was the realization not of a vision but rather of the more prosaic desire to find a formula for those num bers that are described in terms of their ability to satisfY certain kinds of cubic polynomials. Such a number is said to be algebraic ofdegree three and has the property that there is some integer which, when added to some combination of multiples of its first, second, and third powers, gives zero. For example, you might look for a number such that three times its third power when added to minus two gives zero. * One such number is the cube root of two-thirds. Our goal is to arrive at a general scheme which, given the relation that we seek to satisfY, is able to give a formula for a number that satisfies the original cubic polynomial. This is analogous to the quadratic formula that some of us learn in order to solve relations involving at most the second power. Nahin tells of the sixteenth-century mathematicians who posed problems and publicly challenged one another to "duels" for their solution. Through these public displays of intellectual acuity, winners 'I.e., a solution [0 [he equation 3.0
-
2=
o.
Riemann and His "Very Likely" Hypothesis
73
hoped to gain the patronage of the intellectually curious aristocracy, while losers suffered wounds only to the ego. One such mathemati cian-gambIer-duelist, Girolamo Cardano (1 501-1 576), found himself forced to use, or at least manipulate, the square root of a negative number in order to solve a certain cubic polynomial. It was (in his words) "mental torture" to use such "sophistic" numbers. These numbers that so tortured Cardano are instances of complex numbers. Euler was the first to symbolize the square root of minus one as i. Other examples of complex numbers are 3 + 4i, -2 + 3i/2, and 1[i. Most generally, a complex number is expressed as a + ib where a and b are any real numbers. In this case a is the realpart and b the imaginary part of the complex number a + ib. This augmentation of our mathe matical language now enables the creation of a completely self consistent arithmetic for this new and all-encompassing world of numbers. The totality of this "complex universe" of number, with its nesting of the real numbers within the complex numbers, and the rationals within the reals, and the integers within the rationals, and finally, the natural numbers within the integers, is pictured in Figure 9. Despite their daunting name, complex numbers are actually the source of much simplification within mathematics. Problems that appear intractable in a restricted world which uses only real numbers magically decompose when cast into the larger complex setting. Rather than being complex, these are numbers of ease and simplicity, and in their utility for the sciences and engineering they are perhaps even more real than the real numbers. Even in their definition, they are more like life itself: partly what is and partly what we wish for, part real and part imaginary.
The Complex Plane Real numbers live on the one-dimensional number line, but the inde pendent nature of the real and imaginary parts of a complex number suggests a two-dimensional representation, like coordinates in the Cartesian plane. This setting is the complexplane, in which the labels real and imaginary replace the east-west and north-south, or x and y, of the Cartesian plane. A portion of the complex plane is represented in Figure 10 with a
74
STALK I N G T H E RIEMANN H YPOT H E S I S
Complex U niverse Reals Transcendentals
--Ii
I
..
I
.
"
'\ "
I:
/-- -;;/